Design of Water-Splitting Photocatalysts by First Principles Computations by Yabi Wu B.Sc., Peking University (2008) Submitted to the Department of Materials Science and Engineering in Partial Fulfillment of the Requirements for the Degree of MASSACHUSETTS Doctor of Philosophy MASSACHE OF TECHNOLOGY at the MAjY 14 2014 MASACHUSETTS INSTITUTE OF TECHNOLOGY LIBRARIES October 2013 ©2013 Massachusetts Institute of Technology. All rights reserved Signature of Author ............................................................................ Yabi Wu Department of Materials Science and Engineering October 2013 Certified by ................... ....................... Gerbrand Ceder R.P. Simmons Professor of Materials Science and Engineering Thesis Supervisor Accepted by. ........ ... .................... Gerbrand Ceder Chair, Departmental Committee on Graduate Students ..................... E Abstract This thesis focuses on the design of novel inorganic water-splitting photocatalysts for solar applications using first principles computations. Water-splitting photocatalysts are materials that can photo-catalyze the water-splitting reaction under certain conditions. They provide an alternative way to capture and store the energy from the sun. Currently, the energy conversion efficiency of photocatalytic devices under solar illumination and in pure water (pH=7) is still far from the commercialization target. The design of new photocatalysts with better potentials is the key to solve this problem. We have first developed a so-called three-step method to compute the relative position of a semiconductor's conduction band (valence band) vs. the H2/H2 0 (0 2/H 2 0) level in solution from first principles. The merits of the method have been highlighted, and the performance of the method has been tested and compared with the performance of other methods. We conclude that the three-step method provides the desired accuracy for high throughput screening at an acceptable computational cost. We have designed a three-tier first principles high throughput screening system to identify new water-splitting photocatalysts by examining the phase stability, band gap and band edge positions of the candidate compounds. We construct the screening system by integrating the three-step method together with other previously developed methods in our group. We use the system to screen about 3000 different materials. Through the screening, most of the known water-splitting photocatalysts have been reproduced and, more importantly, sixteen new promising candidates have been proposed. Properties of these new candidates have been analyzed and compared to those of the known photocatalysts. Some particularly promising ones are highlighted. Ti3 0 3N2 is one of the identified candidates from the high throughput screening, and is particularly interesting as it has good phase stability, a low band gap and suitable band edge positions. In addition, it has the same crystal structure as Ta 3N5 , which is also a photocatalyst with a low band gap. This leads to our study on the Ta 3N5 :Ti 3O 3N2 solid solution as a water-splitting photocatalyst. Using first principles computations, we study the phase stability, band gap and band edge positions of the solid solution. The results suggest that the Ta3N5 :Ti 3O3N2 solid solution may have a better potential than both its end members as a water-splitting photocatalyst. Thesis Supervisor: Gerbrand Ceder Title: R.P. Simmons Professor of Materials Science and Engineering Acknowledgements I have received a lot of help in my graduate life. I owe many thanks and much appreciation to everyone who helped me throughout graduate school. I would like to take this opportunity to express my gratitude to all of them. I would like to first thank my advisor, Professor Ceder. I am very lucky to have him, a professional, inspiring and intelligent advisor, supervise my Ph.D. research. It is him who led me to the field of water-splitting photocatalysis. Otherwise, I may have not noticed this interesting field. He always encourages us to solve the major problems of a field. Therefore, I chose my Ph.D. project as the design of new water-splitting photocatalysts. He has provided a very stable environment for my research and is always very helpful in generating ideas and discussing research results. In addition, his teaching and presenting style deeply influenced me. I am also deeply grateful to my family members who are always my firmest support. My parents understand me a lot and always give me an indispensable mental support. My wife, being also a MIT graduate student, has accompanied me all the way along my graduate life. I always feel very lucky that I have her around these years, sharing the joys and sadness, taking care of each other, and enjoying the research experience. I want to thank all my collaborators, all members of Ceder group and all my friends too. This thesis would have been impossible without the important contributions of my coworkers, Maria K. Chan, Shyue Ping Ong, Oliviero Andreussi, Predrag Lazic, Geoffroy ~ 3~ Hautier, and Kristin Persson. Meanwhile, many of my research progresses were driven by the valuable discussions with Ruoshi Sun, ShinYoung Kang, Rahul Malik, Aziz Abdellahi, Alexander Urban, and Yifei Mo. In addition, I sincerely thank every member in the Ceder group for creating such a creative and stimulating working atmosphere, and thank all my friends for sharing their life and research experience with me. I have learned a lot from each of you. Table of Contents Abstract.............................................................................................................................. 1 Acknowledgements ........................................................................................................ 3 Table of Contents ....................................................................................................... 5 Chapter I. Introduction................................................................................................. 7 1.1 W ater-splitting photocatalysis as a technique to utilize the solar energy ............ 13 1.2 Oxides, non-oxides, and oxynitrides as water-splitting photocatalysts ............... 15 1.3 Design of new photocatalysts by first principles high throughput screening ......... 18 1.4 Motivation and overview of this thesis ............................................................... 20 Chapter II. Prediction of semiconductor band edge positions in aqueous environments from first principles................................................................................ 23 2 .1 Intro du ction ............................................................................................................. 23 2.2 Methodology development................................................................................. 27 2.3 Computational details and results ....................................................................... 29 2.3.1 Semiconductor bulk computation................................................................. 30 2.3.2 Liquid water bulk computation...................................................................... 31 2.3.3 Semiconductor-water interface computation............................................... 35 2.3.4 Results of CB edge positions relative to water H 2/H 2 0 level........................ 37 2 .4 D iscu ssion ............................................................................................................... 38 2 .5 C on clu sion s ............................................................................................................. 40 Chapter III. First principles high throughput screening of oxynitrides for watersplitting photocatalysts ............................................................................................... 42 3 .1 Introdu ction ............................................................................................................. 42 3.2 Construction of the screening system.................................................................. 43 3.2.0 Generation of the candidates ....................................................................... 44 3.2.1 Phase stability screening............................................................................... 46 3.2.2 Band gap screening........................................................................................ 47 3.2.3 Screening of band edge positions in aqueous environment.......................... 48 3.3 Results of the screening........................................................................................ 50 3.3.1 Binary nitrides ............................................................................................... 50 3.3.2 Ternary oxynitrides......................................................................................... 53 3.3.3 Quaternary oxynitrides ................................................................................. 58 3.4 Discussion ............................................................................................................... 65 3.5 Conclusions ............................................................................................................. 71 Chapter IV. First principles study on Ta 3N5 :Ti3O 3N 2 solid solution as a watersplitting photocatalyst.................................................................................................. 73 4.1 Introduction ............................................................................................................. 73 4.2 M ethods................................................................................................................... 76 4.3 Results ..................................................................................................................... 78 4.4 Discussion ............................................................................................................... 81 4.5 Conclusions ............................................................................................................. 85 Chapter V. Conclusions and future work.................................................................. 87 Bibliography .................................................................................................................... 90 Chapter I. Introduction Step (3) H 2 H+ Step (3) Cocatalyst .-- 2 nanoparticle Step (2) Step (1) e- H2 h+ Recombination, Particulate hv > E9 photocatalyst Figure 1.1 A schematic illustration of the water-splitting photocatalysis*. Step (1): photon absorption and electron-hole excitation. Step (2): electron-hole separation and transportation. Step (3): HER and OER activation. hv-1.23eV H210(l) -* 1 H2 (9) +'-0 2 (9) (AG = 237kJ/mol) (1.1) Water-splitting photocatalysts are materials that can photo-catalyze the water-splitting reaction (Eqn. 1.1) under certain conditions. Although they can be in the form of either inorganic semiconductors[1-3] or molecular materials[4-8], we focus on inorganic photocatalysts in this thesis. In the photocatalysis process (Fig. 1.1), the photocatalyst absorbs photons from artificial or natural lights, and generates excited electron-hole pairs. After separate with each other, the electrons and holes travel to the photocatalyst-solution interface and activate the hydrogen evolution reaction (HER) and oxygen evolution * Reprinted (adapted) with permission from J.Phys. Chem. C, 2007, 111 (22), 7851-7861. Copyright (2007) American Chemical Society. reaction (OER) respectively[ 1-2]. The difference of these two reduction potentials is 1.23 eV, indicating that the band gap of the photocatalyst must be greater than this value otherwise the absorbed photons do not have enough energy to drive the photocatalysis process[1-2]. In many cases, some co-catalysts are also used to reduce the kinetic barrier of the HER or OER so that the reaction rate can be improved[1-3]. Potential / V (NHE) + 0.0 V ------------------------------------- +1.23V I---------------------------------- (a) (b) H2/H 20 02/H20 (c) Figure 1.2 A diagram of the energetic requirements for the band edge positions at the photocatalyst-solution interface. Blue lines represent the CB positions while red lines represent the VB positions. (a) favorable band level arrangement; (b) unfavorable VB position; (c) unfavorable CB position. Besides a band gap greater than 1.23 eV, the photocatalyst must also have its conduction band (CB) higher than the H2/H 20 redox level in solution and have its valence band (VB) lower than the 0 2/H2 0 redox level in solution[9-1 1]. Otherwise, the HER or OER is not energetically favorable (Fig 1.2). Here and throughout this thesis, "higher" always refers to more negative in the Normal Hydrogen Electrode (NHE) reference while "lower" always refers to more positive in the NHE reference. It is worth to note that the CB and VB positions considered here should be the band edge positions at the photocatalyst~8 ~ solution interface. They are usually different than the band edge positions in the bulk photocatalysts due to the band bending effect[1 1]. We will explain this point further in Chapter II. Potential /V (NHE) H2 H20 + 0.0 V Ox + 1.23 V H2/H 20 Intermediate redox level Red ------------------------- n /H 2 I 2 H20 $ 02 Figure 1.3 A schematic illustration of the z-scheme photocatalysts. Blue lines represent the CB positions while red lines represent the VB positions. Alternatively, two materials may work together in a so-called z-scheme[1-2, 12-14] (Fig 1.3), one for anode where the OER takes -place and one for cathode where the HER takes place. In this case, two photons are absorbed in order to provide one effective pair of electron and hole, which are the electron from the cathode and the hole from the anode. To cycle the other pair of the excited electron and hole (the electron from the anode and the hole from the cathode) which is not effective in the water-splitting reaction, some intermediate redox level should be introduced (Fig. 1.3). This intermediate redox level is often provided by certain ion couples in the solution[1-2, 12-14]. In the z-scheme, the photocatalysis process is energetically favorable if the anode material has its VB lower than the 0 2 /H2 0 level and its CB higher than the intermediate redox level, while the cathode material has its CB higher than the H2/H2 0 level and its VB lower than the intermediate redox level [1-2, 14]. In principle, the z-scheme based photocatalysts have advantages over the single material photocatalysts in solar applications, since they are more flexible in band gaps and may absorb more faction of the solar spectrum[14], although they use only half of the excited electrons and holes. However, the z-scheme design dramatically increases the complexity of the device and the fabrication costs. In addition, it is usually not easy to find suitable ion couples that can provide the desired intermediate redox level. Therefore, we still mainly focus on the single material photocatalysts in this thesis. There are mainly two types of water-splitting photocatalytic devices (Fig. 1.4), the photoelectrochemical (PEC) cell[1-2] and the so-called powdered device[14]. In a PEC cell device, the photocatalyst forms one of the two electronodes (e.g. a photo-anode in Fig. 1.4) while the other electronode is formed by a metal (e.g. platinum). The two electronodes are wired, so that the electrons and holes which are generated at one electronode, can travel to the cathode and anode, and activate the HER and OER respectively. In addition, a bias voltage can be applied to this device between the cathode and the anode. The bias voltage can shift the band edge positions of the photocatalyst relative to the water redox levels in the solution. By this way, materials with a band gap larger than 1.23 eV but unfavorable CB or VB positions may still achieve photocatalytic ~10 ~ performances. However, applying a bias voltage consumes energy and thus reduces the efficiency of the device. Therefore, a large bias voltage should be avoided. For this reason, photocatalysts are still requited to have its CB and VB position close enough to the H2/H20 and 0 2/H20 level even a bias voltage can be applied. In a powdered device, the photocatalyst in the powdered form are directly put into the solution. By some mechanisms of electron-hole separation such as p-n junctions[ 15-19], built-in electric fields[20-22], and so on, the exited electrons and holes can travel to different places at the photocatalyst-solution interface, where they activate the HER and OER respectively. The powdered device is simpler than the PEC cell device, and it is easy to be scaled up as it is not wired. Therefore, the powdered device has a good potential in practice. The disadvantage of this device is that no bias voltage can be applied on it, indicating that the requirement of the band edge position in Fig 1.2 should be strictly satisfied. In addition, unlike the PEC cell, the powdered device generates H2 and 02 at the same particles, so it requires a gas separation step which is non-trivial. ~ 11 ~ Vbias -MMM II t A> 4 pV H2/H 20 0 2/H 2 0 (a) A powdered photocatalysts H2 H20 moo *0 O H20 '%~IE a + 02 H21.2 0. (b) Figure 1.4 A schematic illustration of two types of photocatalytic devices. (a) PEC cell device; (b) powdered device. 1.1 Water-splitting photocatalysis as a technique to utilize the solar energy Although the technique of photocatalysis has many interesting applications, such as water purification[23-26], artificial photo-synthesis[5-6, 12, 14, 21], photoelectrochemical CO 2 reduction[4, 8], and so on, the major interest in this thesis is to apply it to capture and store the solar energy by photo-catalyzing the water-splitting reaction[ 1-3, 14]. Indeed, at a power level of 1 000W/m 2 , the solar energy incident on the earth's surface by far exceeds all human energy needs[l, 27]. Photovoltaic[28] and electrochemical solar cells[9, 28-29] that convert solar energy into electricity can reach up to 55% ~ 77% efficiency[30-32] but remain uneconomical because of high fabrication costs, insufficient light absorption[33] and inefficient charge transfer[28]. Photocatalysis, which directly converts solar energy into H2-based chemical energy, may serve as an alternative way to utilize the solar spectrum[ 1-3, 14]. Currently, the primary objective in this field is to find photocatalyst materials which can achieve above ~10% energy conversion efficiency (ECE) without any bias voltage[1]. This is the performance that the US Department of Energy considers to be the minimum for commercialization[34]. Since the discovery of the first photocatalytic water-splitting system based on TiO 2 and Pt in 1972 by Fujishima and Honda[35-36], more than 130 inorganic materials have been demonstrated to exhibit photocatalytic performance for water-splitting[1]. However, under solar illumination, the efficiency of current photocatalytic devices is still well below the commercially viable level[1]. So far, the GaN:ZnO solid solution achieves the ~ 13~ highest efficiency of any photocatalyst in pure water (pH = 7) and under visible light, with a quantum efficiency -2.5%[1]. There are two major reasons for the low efficiency. One reason is the lack of low band gap photocatalysts. It is commonly known that the optimal range of band gap for absorbing the solar spectrum is 1.1 eV ~ 1.7 eV[37]. After considering multiple sources of efficiency loss in realistic conditions such as reflection loss, quantum-yield loss, absorption loss and collection loss, Refs. [38-39] concluded that a band gap around or less than 2.0 eV is necessary for a single material water-splitting system to reach a ECE level above ~10% under AM 1.5 solar radiation. However, not many photocatalysts with a band gap around this level have been found. To solve this problem, new photocatalysts with a low band gap need to be identified. This is the major objective of this thesis. The other reason is that the reaction rate of current water-splitting photocatalytic device is too low. The water-splitting reaction has very slow kinetics, as it is a process involving four electron/holes[1-2, 40]. The time from the generation of the electron-hole pairs to the activation of the HER and OER is significantly longer than the typical electron-hole lifetime in semiconductors[40-41]. Therefore, a large fraction of electrons and holes recombined with each other before going into the solution. To solve this problem, kinetic barrier of the water-splitting reaction at the interface should be reduced and the electronhole lifetime should be enhanced. Besides to improve the photocatalysts, such as finding photocatalyst materials with better kinetic and transportation properties, growing purer compounds, increasing surface trapping states[42], forming p-n junctions[ 15-19], introducing innate electric fields[20-22] and so on, this purpose may be also achieved by ~14~ searching and using good co-catalysts. Co-catalysts can often significantly reduce the reaction barrier and improve the water-splitting reaction rate. For instance, the rate of H2 evolution reaction becomes ~100 times faster when 5 wt% Ru is added to TaON [43], and in the ZnO:GaN solid solution, both H2 and 02 evolution changed from negligible to clearly observable when 5 wt% RuO 2 was present[44]. It is noteworthy that the performance of co-catalysts is system dependent, and finding good co-catalysts for individual photocatalysts is an active research topic in this field as well, though it is not the main focus in this thesis. 1.2 Oxides, non-oxides, and oxynitrides as water-splitting photocatalysts - --... - --.... --... ... .... .... ... .... --..--- --... - .... - -- - --...--- tj ---. .0 1 0 0 - ---..- ..---.---- ..------...--US04 La:NaTaO3 TiO 2 C 1I - M2 La2Ti, C, o CdS Ge3N KBa M 2 0 (M=K,Rb,Cs) TiO2 (Anatase) Target NbTO2 La4CaTi0 7 (M=KRb) HCa2Nb3O,1 Zn:La203/Ga0 W E Sr2Nb, MNbO K2Sr, Ta0 + a,4 GaN:ZnO MZnS (M=Ni,Pb,Cu) (M=SrBaS.) Z [2 n Zn:CIn(OH),S, jN:inTaO BiV4 2 4 4-+BV 10 LaTiO2N UV 0.1 200 Cr/Ta:Sr 300 0, VISIBLE I-+ wo 400 500 600 Wavelength (nm) Figure 1.5 Quantum efficiency and band gap of some oxide photocatalysts*. More than 130 inorganic materials have been demonstrated to exhibit photocatalytic performance for water-splitting over the last three decades, and over 80% of them are oxides[1]. This is primarily because oxides are the most well-studied chemistries, and * Reprinted (adapted) from a website http://payneresearch.org/research/photo-electrochemical-pec-water-splitting/ ~ 15 ~ many of them are stable in aqueous environment. Some of these oxide photocatalysts are summarized in Fig. 1.5. The figure shows two major trends of these oxides: (1) most of them contain do or d'0 cations; (2) most of them have a band gap larger than 2.7 eV(~ 450 nm), which is too high for achieving the commercially viable ECE level under solar illumination. The primary reason for the large band gap of these oxide photocatalysts is that their VB is typically dominated by 0 2p states which are about 3 eV lower than the H2/H 20 level[3]. This gives a lower bound roughly 3 eV to the band gap of the oxide photocatalysts, since the CB of photocatalysts needs to be above or close to the H2/H20 level. As a result, to achieve a lower band gap, the VB position of the oxide photocatalysts must be raised. There are mainly two strategies to reduce the band gap of these oxide photocatalysts. The first strategy is to dope other cations or anions into the oxides. Different cations and anions have been studied by first principles computations as possible dopants into the photacatalytic system. For instance, cations (V[45], Mn[46], Fe[46], Co[47] and so on) or anions (C[48], N[49-50], P[51], S[51] and so on) doping and some co-doping strategy (doping cations and anions simultaneously)[52-54] have been studied by Density Functional Theory (DFT)[55-56] for TiO 2 , the earliest experimentally demonstrated photocatalyst[35-36]. Similar studies have also been performed for other photocatalysts such as ZnO[57-58], W03[59], SrTiO 3 [60-61] and so on. To generally summarize these studies, cation doping exhibits very limited band gap reduction, while anion doping or co-doping sometimes significantly reduce the band gap. This is because that the cation doping usually has little effect on the VB position of the oxide materials while the anion ~16 ~ doping may sometimes raise the VB position significantly. However, doping anions into oxides are very challenging in practice. The other strategy is to form solid solutions. Due to band bowing effects[62-63], solid solutions can have a lower band gap than both of their end members. For example, the GaN:ZnO solid solution has a band gap of 2.4 ~ 2.8 eV, while both GaN and ZnO have a band gap larger than 3.0 eV[3]. Following this strategy, one may form solid solutions of two or more photocatalysts with a same crystal structure and obtain new photocatalysts which have a lower band gap than both of their end members. More detail of this point will be presented in Chapter IV. Besides to improve the oxide photocatalysts, one may also find photocatalysts with a low band gap in other chemistries. For instance, nitrides and sulfides may have a higher VB compared to oxides, since their VB are typically dominated by N2p and S3p states, which are higher in energy than the O2p states. In fact, a few nitride compounds such as Ta 3N5 [64-65], GaN[44], Ge 3N4[66] and so on have been experimentally demonstrated as water-splitting photocatalysts. Some of them indeed have a low band gap (e.g. Ta 3N5 , ~2.1 eV[64-65]). The potential problem of these materials is that they are usually less stable than oxides in aqueous environment[3]. This limits their commercial exploitation as water-splitting photocatalysts. As a balance of the band gap and the aqueous stability, oxynitrides have been recently proposed as a promising chemical space for low band gap photocatalysts[3]. They usually have a VB dominated by a mixture of 0 2p and N2 p states, which is typically higher in ~ 17 ~ energy than the VB of pure oxides[3]. In the meantime, they are often more stable in aqueous environment compared to pure nitrides[3]. Following this idea, experimentalists have identified a few oxynitrides compounds as water-splitting photocatalysts with a low band gap, such as TaON (2.5eV)[3], SrTaO 2N (2.1 eV)[67], BaTaO 2N (1.9 eV)[67], CaTaO 2N (2.4 eV) LaTiO 2N (2.0 eV)[68], CaO.25 LaO. 7 5TiO 2 .2 5NO. 7 5 (2.0 eV)[68], LaTaO 2N (2.0 eV)[69] and CaNbO 2N (1.9 eV)[69]. In this thesis, we use first principles computational approaches such as high throughput screening[70-73] to identify new oxynitride compounds as promising photocatalysts. 1.3 Design of new photocatalysts by first principles high throughput screening Although theorists and experimentalists have put a lot of efforts in this field and have harvested fruitful results, the ECE of the currently known photocatalysts are still well below the commercially viable level (~10%)[1]. This spurs us to continue to identify new photocatalysts with better potentials by both experimental and computational approaches. High throughput computational screening whereby one computationally assesses key properties of a large number of compounds for a given application has shown its merit in many fields, such as the design of new battery materials[74-77], thermoelectric materials[78-79], piezoelectric materials[80], and organic photovoltaic materials[81-83]. The development of ab-initio property prediction methods and their automation makes it possible to examine thousands of material candidates for a few desired properties[84-85]. ~18 ~ In order to apply this technique to identify new water-splitting photocatalysts, one should first determine which properties are the identifiers for photocatalyst materials. We have mentioned earlier that water-splitting photocatalysts for solar applications should have the following properties[1-2]: (1) they should be thermodynamically stable enough so that they can be synthesized; (2) they should have a band gap that suitable for absorbing solar spectrum; (3) their CB should be higher than the H2/H 20 level while their VB should be lower than the 0 2/H2 0 level (this requirement can be slightly loosen in a PEC cell device which allows to apply a bias voltage); (4) they should be stable in aqueous environment; (5) they should have fast and selective kinetics for water-splitting reaction; (6) they should have good transportation properties and a long electron-hole lifetime; (7) they should be inexpensive to manufacture. Apparently it is not practical to consider all these properties in the high throughput computational screening. Computational predictions on some properties, such as kinetics properties and transportation properties are still too computationally expensive and not stable enough to be handled in a high throughput manner. In fact, we have considered only the first three properties (i.e. the phase stability, the band gap, and the band edge positions in solution) in the high throughput screening in this thesis. The methods and the results of the screening will be explained in detail in Chapter II and Chapter III. For other properties, it may be better to investigate them in individual systems after new candidates have been identified from the high throughput screening. ~19 ~ 1.4 Motivation and overview of this thesis Material problem is the core problem in the field of photocatalysis and is the key to achieve high-efficiency water-splitting photocatalysis under solar illumination. To solve this problem, new photocatalyst materials with better properties need to be designed. In this thesis, we present our work on designing new water-splitting photocatalysts by first principles computations. We mainly use the high throughput computational screening technique and focus on the three key properties of photocatalysts, the phase stability, the band gap, and the band edge positions in solution. The computational methods that are used in the high throughput screening system to predict these properties should be accurate, stable, and computationally inexpensive. For predicting the phase stability and the band gap, such methods are available as the convex hull construction method[86] and the A-sol method[87] which will be further introduced in Chapter III. However, methods that predict the CB and VB position of a semiconductor vs. the water redox levels are not satisfactory in either accuracy or computational cost. To solve this problem, we have developed a first principles computational method, the so-called three-step method[ 11]. It can predict the relative band edge positions of a semiconductor (i.e. the distance of the CB vs. the H2/H2 0 level and the distance of the VB vs. the 0 2 /H2 0 level) with an acceptable computational cost. More importantly, the method takes the effect of the aqueous environment into account and thus significantly improved the accuracy of the results[ 11]. We will explain the method in detail in Chapter II. ~20~ By integrating the convex hull construction method[86], the A-sol method[87] and the three-step method[ 11], we construct the high throughput screening system for identifying new water-splitting photocatalysts. We implement this screening system mainly on oxynitride chemical space, as the reason we mentioned in Section 1.2. We have screened 2948 different candidate compounds including 68 binary nitrides, 1503 ternary oxynitrides and 1377 quaternary oxynitrides. The screening has reproduced most of the known water-splitting photocatalysts and has also found sixteen new promising candidates. Both the algorithms and the results of the screening will be presented in detail in Chapter III. Ti6O 3N 2 , one of the identified candidates from the high-throughput screening, is particularly interesting. Computational results indicate that it can likely be synthesized[71]. In fact, it is recently declared to have been synthesized from a website[88]. Its CB and VB are predicted to be bracketing the water redox levels[71], indicating that it may activate both the H2 and 02 evolution reaction without any bias voltage. Its band gap is predicted to be 2.37 eV, which is in the visible light region. In addition, we find that it has the same crystal structure as Ta 3N5 , which is also a photocatalysts with a low band gap. This inspires us to study the solid solution of these two materials, because that by forming the solid solution, we may obtain a material with a lower band gap than both of the end members. Using first principles computations, we have shown that the solid solution of Ta 3N5 and Ti3 0 3N2 can likely be synthesized and remain stable. The lowest band gap of the solid solution is predicted to be around 2.0 eV at a composition around 50%:50%. Therefore, a band gap reduction of ~0.2 eV is ~21 ~~ achieved by forming the Ta 3N 5 :Ti 3O 3N 2 solid solution. More details of this work will be provided in Chapter IV. ~22~ Chapter II. Prediction of semiconductor band edge positions in aqueous environments from first principles In this chapter, we present a novel first principles method, the so-called three-step method[11], which computes the CB position of semiconductors relative to the water H2/H 20 level using DFT[55-56] with semi-local functional[89-90] and classical molecular dynamics (MD). We test the method on some photocatalyst materials which have their band edge positions measured in experiments. The predicted band edge positions are within 0.34 eV of the experimental data, with a mean absolute error of 0.19 eV[1 1]. Both the accuracy and the computational cost of the method are acceptable for the application of the first principles high throughput screening. Therefore, we integrate this method into the high throughput screening system in Chapter III. 2.1 Introduction As was shown in Fig. 1.2, one crucial requirement for a water-splitting photocatalyst material is that its CB should be higher than the H2/H2 0 level of water and its VB should be lower than the 02/H 2 0 level. This requirement ensures that the water-splitting reaction is energetically favorable without a bias voltage. Therefore, the knowledge of a semiconductor's CB and VB band edge positions, relative to the H2/H 20 level and the 0 2 /H2 0 level in solution respectively, is important for the design of a water-splitting photocatalyst[9- 11]. An ab initio approach to obtain such band edge positions is preferred ~23~ as it can be used as a scalable approach to investigate a large number of possible materials. A straightforward attempt for this purpose is to compute both band edge positions of semiconductors and water redox levels, relate them to a common reference, and then calculate their difference. The vacuum level is a natural candidate for the common reference. The band levels of semiconductors and the water redox levels relative to the vacuum level have been respectively computed using DFT in Refs. [91-92]. However, the problem comes from the fact that the band realignment at a semiconductor-water interface is not equal to the difference between the band realignment at the semiconductor-vacuum and water-vacuum surfaces. This difficulty is explained in Ref. [93] for the metal-semiconductor interfacial system. The main reason is that the dipole at metal-semiconductor interface is not equal to the difference between the surface dipoles at the metal-vacuum and semiconductor-vacuum surfaces. For the semiconductor-water interfacial system, we will show later in the Section 2.4 that the error due to this problem is up to 0.7 eV. Apart from this approach, a few other computational methods have also been proposed in the literature. In Ref. [94], hydrogen levels in semiconductors and insulators have been aligned by a valence-band offset method[95-96]. This method assigns the absolute energy scale by setting the CB of Si to the measured EA of Si, calculates the formation energies of interstitial He/H/H- species, and sets the Fermi-level position at the energy level at which the positive (with H*) and negative (with H-) charged states share the same energy as the hydrogen levels within the semiconductors and insulators. However, this method ~24~ also avoids directly dealing with a semiconductor-water interface system and thus may have similar band alignment problems as the vacuum reference method. The mean absolute error of the method is around 0.4 eV. A method which directly deals with the semiconductor-water system was developed in Ref. [97]. In the paper, the authors computed the band edge positions of TiO 2 relative to water redox levels, using the generalized gradient approximation (GGA)[90] and ab initio MD. In principle, the method can be generalized to compute the band edge positions of other inorganic semiconductors too. However, the errors for TiO 2 's CB and VB positions found by the method were substantial, at respectively 0.4 eV and 1.6 eV[97]. They argue that the error may come from the simplified assumption that the zero-point energy (ZPE) of a proton in a solvated H3 0 ion can be approximated by the net ZPE of a dummy proton in an isolated pseudo H30 molecule, a molecule with the same atomic configuration as an isolated H30 ion but with neutral charge[98]. Since the ZPE is directly added to their results and is as large as 0.5 eV, the assumption may introduce significant errors. Compared to the methods mentioned above, the first principles method presented in this chapter, which computes the CB position of a semiconductor relative to the H2/H2 0 level in solution, has the following advantages: (1) it is applicable for general inorganic semiconductors; (2) it directly deals with band realignment effects introduced by the semiconductor-water interface; (3) it is mainly based on total energy calculations using DFT-GGA, with reasonably low computation cost. In fact, an approach for the computation of band edge alignments across a solid-solid interface has previously been developed. The band alignment between two semiconductors[99-100], and the Schottky barrier heights between a semiconductor and a metal[101], are typically computed with DFT by three sub calculations, two bulk calculations to compute the difference between the target energy level (CB, VB or Fermi energy) and the average Hartree potential of each solid, and an interfacial slab computation to compute the Hartree potential difference between the two solids. There are several challenges when replacing one solid system by liquid water. Since liquid water lacks periodicity, and ab initio MD can produce considerable errors for water[102-103], it is non-trivial to construct a cell with accurately representative atomic configurations of liquid water in DFT. Instead, we use the idea proposed in Ref. [104] and equilibrate a classical MD computation of water at room temperature. Snapshots of the water configuration at different MD time points are then computed with DFT. By combining the band alignment method for solid-solid system and the idea of using snapshots of classical MD water configurations for DFT, we develop a so-called threestep method for computing CB band edge position relative to the H2/H 20 level. In the next sections, we introduce our methodology in detail, and present the computational results obtained with this approach for six common water-splitting photocatalyst materials, TiO 2 , W0 3 , CdS, ZnSe, GaAs and GaP. Finally, we computational results to experimental data. ~26~ compare the 2.2 Methodology development Semiconductor Solution BebulkEc Ecedge*- AedgeA sol edge H sol bulk semic k Hs Hssemibedg bl Figure 2.1 A schematic diagram of the band alignment at the semiconductor-water interface. Ecbulk = CB in the bulk of the semiconductor; Ecedge = CB at semiconductorsolution interface; Abulk = Acceptor level (H 2/H 2 0 level of liquid water in this work) in the bulk of the solution; Aedge = Acceptor level at the semiconductor-solution interface; Hsemi bulk = Hartree potential in the semiconductor bulk; Hsemi edge = Hartree potential on the semiconductor side at the semiconductor-solution interface; Hso _bulk = Hartree potential in the bulk of the solution; Hsol edge = Hartree potential on the solution side at the semiconductor-solution interface Ece Aedge - Hsem = Ecb - H em -Hsol _eedge =Abulk - Hsol (2.1) (2.2) bulk Ecedge - Aedge = (Ecedge - Hsem_ edge = (Ecbulk - Hsemi bulk) (Age - - (Abulk - H 01 edge) + (Hsemi edge - Hol_ edge) (2.3) bulk) + (Hsemiedge - Hsol edge) Fig. 2.1 shows a schematic diagram of band alignments at an interface, and introduces the terminology we will be using. Our objective is to compute the CB band edge position relative to the solution acceptor level (H 2/H 2 0 level of liquid water in this work) at the interface, i.e. Ecedge - Aecge . We assume that the band alignment is due to electrostatic effects (electrons and ions redistribution near the interface due to Fermi energy ~27~ realignment). So the energy levels and Hartree potential change by the same amount everywhere in space and their difference remains unchanged. Thus, we obtain Eqn. 2.1 and Eqn. 2.2. Therefore, the term Ecedge -Aedge Eqn. 2.3 indicates that the term Ecedge method. Step 1: compute the term can be computed by Eqn. 2.3. can be obtained by the following three-step Aedge Ecbulk - Hemibulk' i.e. the eigenvalue of the lowest unoccupied eigenstate relative to the average Hartree potential, in a bulk semiconductor system. Step 2: compute the term Abulk - H,0 _bulk'5 i.e. the eigenvalue of the molecular acceptor level relative to the average Hartree potential, in a bulk liquid water system. This is non-trivial and we adopt the idea of using MD atomic configurations for DFT. More details will be introduced in Section 2.3. Hsemi edge Hol edge Step 3: compute the term , i.e. the difference in average Hartree potentials between the semiconductor vs. the liquid water, in a semiconductor-water interfacial slab system. During step 3, we join the bulk cells that we compute in steps 1 and 2 and make a supercell which contains the interface. In this supercell, we compute the variation of the Hartree potential with position. By averaging the Hartree potential on both the semiconductor side and liquid water side, we calculate Hsemi edge - Hsol edge This method has two key features. One is that it includes the band realignment effect yet avoids a large supercell computation. The band realignment effect at the semiconductorwater interface is important for computing the relative energy levels. However, it usually occurs over a distance of 100A to several micro meters from the interface[104]. As a consequence, directly computing the band alignment effect, i.e. the term ~28~ Ecedge - Ecbulk (or Aedge -Abulk ), in a single slab computation is not applicable, since it requires a prohibitively large supercell to converge both Ecedg, (Aedge ) and ECulk (Abulk) in the same system. On the other hand, in the three-step method, the three objective terms, , Ecbulk -H Abulk emibulk -Hsot -bulk , and Hsemi edge -Hol edge are either pure bulk properties or pure interface properties, so a large supercell is not required. In this approach, the band effect is captured by the computation of Hol ege .And the longest dimension of the supercell required to converge Hemi edge Hsemi edge -Hsoledge - realignment to 0.1 eV is typically 30A to 40A. The other important feature is that the three-step method only requires the Hartree potential in the interfacial slab computation but not any energy eigenvalues. This prevents the complicated problem of trying to assign electronic states to specific real space domains of the supercell. 2.3 Computational details and results To test our approach, we select six popular photocatalyst materials: TiO 2 , W0 3 , CdS, ZnSe, GaAs and GaP. The details of their crystal structures are listed in Table 2.1. We applied the method described in Section 2.2 on these materials to compute their CB position relative to the H2/H2 0 level in liquid water, Ecedge - Aedge The computational results are compared to experimental data obtained from Refs. [105-106]. All DFT computations[55-56] are performed with projector augmented wave (PAW)[89] potentials using the plane-wave code Vienna Ab-initio Simulation Package (VASP)[107108]. We use the Perdew-Burke-Ernzerhof (PBE)[90] GGA exchange-correlation functional unless specified otherwise. ~29~ semiconductor TiO 2 Crystal type Rutile (tetragonal) CdS ZnSe GaAs GaP Zincblende Zincblende Zincblende Tetragonal Wurtzite (hexagonal) (cubic) (cubic) (cubic) 136 113 186 216 216 216 P42/mnm P421m P63mc F43m F43m F43m Initial lattice a=4.598 a=7.616 a=4.137 parameters b=4.598 b=7.616 b=4.137 a=5.670 a=5.654 a=5.447 (A) c=2.956 c=3.960 c=6.714 Space group number Space group W0 3 name Table 2.1 Crystal structure information for test materials. 2.3.1 Semiconductor bulk computation To implement step 1 in Section 2.2, we compute the bulk CB relative to average Hartree potential for each selected material in this section. For every material, we optimize the volume, cell shape and atomic positions of the unit cell with a Monkhorst-Pack[109] 6x6x6 k-point grid and plane wave energy cutoff of 500 eV. On the optimized structures, we perform static DFT computations using a fine F-centered l0xl0x10 k-point grid to compute the CB. We also plot the Hartree potential and determine a macroscopic average over the unit cell for every material. The resulting average Hartree potential Hem, buk is zero. This is consistent with the fact that the absolute Hartree potential in an infinite periodic system is customarily set to zero in DFT codes including VASP. The results of Ecbulk -H,,m, bulk are show in Table 2.2. Testing semiconductor Ecbulk -H semi bulk(eV) Table 2.2 Values of Eculk - TiO 2 W0 3.77 1.89 3 CdS ZnSe GaAs GaP 2.91 3.25 3.64 4.08 Hsemi bulk from semiconductor bulk computations. 2.3.2 Liquid water bulk computation Step 2 in Section 2.2 consists of determining the H2/H 2 0 acceptor level relative to the Hartree potential in bulk liquid water. To prepare the water atomic configurations in DFT, we perform a classical MD computation by DLPOLY[110] and use the TIP4P[111] potential to describe the interaction between water molecules. A water system of 128 H2 0 molecules is initially equilibrated at 300K with a relaxed cell size of 18A x 15.6A x 14.6A. At the same temperature, we further perform an NVT MD simulation for 100 ps and take snapshots of the atomic configurations of this TIP4P water system at t = 50ps and t = 00ps. We construct two DFT cells using these two configurations. Before proceeding, we perform two tests to verify that atomic configurations from classical MD produce consistent results in terms of DFT electronic structures. Only the F k-point is used in the DFT calculations of liquid water cells. First, we compute the band gap and plot in Figs. 3a and 3b the density of state (DOS) by DFT using each of the two cells obtained at different MD time points without any further DFT ionic relaxations. The similar band gap values (3.76 eV and 3.89 eV) and similar DOS plots between Fig. 2.2a and Fig. 2.2b indicate that the atomic configurations taken from different time points of classical MD give little difference in the DFT electronic structures. Second, we repeat the process but with full DFT ionic relaxations (cell volume, cell shape, and atomic positions) -31 ~ for the t = 100ps configuration, and the resulting DOS is shown in Fig. 2.2c. The identical band gap values and similar DOS plots between Fig. 2.2b and Fig. 2.2c indicate that DFT ionic relaxations do not alter the electronic structures after the liquid water system reaches equilibrium in classical MD. In addition, all DOS plots in Figs. 2.2a, 2.2b, and 2.2c are very similar to the DOS plots of liquid water in Ref. [104], which implies that the F k-point alone is sufficient to give results consistent with previous work. (a) a.) cj~ 0 a.) ~0 S z Ni -10 I -8 I I I -6 -4 -2 E-E (eV) . 0 I Im 2 4 (b) c(-) -10 -8 -6 -4 -2 E-E, (eV) 0 2 4 -10 -8 -6 -4 -2 E-E, (eV) 0 2 4 Figure 2.2 Total DOS plots for (a) a 128 H20 molecules liquid system with MD atomic configurations at t = 50 ps without DFT relaxation, with a band gap of 3.76 eV; (b) a 128 H20 molecules liquid system with MD atomic configurations at t = 100 ps without DFT relaxation, with a band gap of 3.89 eV; (c) a 128 H20 Molecules liquid system with MD atomic configurations at t = 100 ps with DFT relaxations, with a band gap of 3.89 eV. E, in the x-axis labels is the VB energy eigenvalue. -10 -8 -6 -2 -4 E-E, (eV) 0 2 4 Figure 2.3 DOS plots for a 127H20 + H30+ liquid system. The black solid line is the total DOS while the red dashed line is the projected DOS from the H30+ ion in this system. We enlarged the H30+ DOS 30 times to make it visible on the scale of the total DOS. The DOS peak at approximately 2.0 eV represents the LUMO of the system contributed by the H30+ ion. Ev in the x-axis labels is the VB energy eigenvalue. In order to compute the term Ab.,, - H,,, -bulk, we need to compute the lowest unoccupied molecular orbit (LUMO) level of water because this level is recognized as the acceptor level of water. While the acceptor is nominally the proton (H*), in an aqueous environment the H+ is solvated in multiple H+(H20). configurations[ 112]. The hydronium ion H30+, being the simplest, is especially important for computing the acceptor level in water system. We simulated the hydronium ion in water by fully relaxing an isolated H30+ ion in DFT and then replacing one of the 128 H20 molecules in the liquid water system with this H30+ ion. The 0 atom of the H30+ is placed in exactly the same position as the 0 atom of the replaced H20 molecule. The orientation of the added H30+ ion is randomized. We perform further DFT relaxation for this added H30+ ion in order to optimize the atomic positions and orientation in the water system. A static DFT computation then follows to compute the energy levels of this 127H 20 + H3 0 system. The DOS plot of such a system is shown in Fig. 2.3, which indicates that a level attributed to H30 is indeed the LUMO. We repeat the above process several times but replace a different H 2 0 molecule with H 30 to ensure that our results are not affected by the positions of the H30 ions in the system. The results are shown in Table 2.3. Replaced H2 0 molecule Abulk - H_ ,bulk (eV) Total energy (eV) Table 2.3 Values of Abulk - 1 2 3 4 -0.70 -0.65 -0.62 -0.75 -1788.2 -1788.1 -1787.5 -1787.7 Hso-bulk from liquid water bulk computations. Table 2.3 indicates that the fluctuation in Abulk - Ho,_bulk in the cell is less than 0.1 eV. We will use Abulk - due to the position of H30ion H,0 1 _bulk = -0.70e V in subsequent calculation since it corresponds to the lowest total energy among all four systems. 2.3.3 Semiconductor-water interface computation This section describes how the semiconductor-water interface calculation (step 3 in Section 2.2) is implemented. We aim to compute the Hartree potential difference between the semiconductor bulk cell and the liquid water cell in an interfacial slab system. The interfacial cell is constructed by joining several layers of the semiconductor bulk cells in Section 2.3.1 and the liquid water cell in Section 2.3.2 together. For each semiconductor, we perform a convergence test in that we increase the number of layers of semiconductor cells until the Hartree potential difference between the semiconductor side and liquid water side is converged to 0.1 eV. The converged Hartree potential profile along the slab direction for TiO 2 is shown in Fig. 2.4 as an example. The calculated value of Hem Hsol edge for each test compound is listed in Table 2.4. Only the F k-point is used for these DFT computations. 4 2 - - --- - .-- . - --- - -- 0 H so'_edge=2.48 eV U. U-- -- dgeU Hsm_edge =-2.00 eV _4 0 5 10 15 ()20 z(A) 25 30 34.4 Figure 2.4 Calculated Hartree potential profile of a stoichiometric TiO 2 -water slab system. The vertical green dashed line indicates the interface. The left side is semiconductor TiO 2 and right side is water. The black solid line is the planar-averaged Hartree potential as a function of cell dimension normal to the interface. The red dashed line with square makers indicates the planar-averaged Hartree potential of TiO 2, Hsemi edge. The blue dashed line with circular makers indicates the planar-averaged Hartree potential of liquid water, Hs0 i edge. Testing semiconductor TiO 2 W0 3 CdS ZnSe GaAs GaP Hsemi edge -2.00 -1.11 -1.02 -1.05 -1.41 -1.56 Hso_ edge 2.48 2.02 1.33 1.30 1.86 1.93 -4.48 -3.13 -2.35 -2.35 -3.27 -3.49 Hemi Hsol _edge edge - Table 2.4 Values of Hseiedge - HsOjedge from interfacial slab computations. 2.3.4 Results of CB edge positions relative to water H2/H2 0 level By substituting the terms Eculk -- H , A,,, and Hem, egH- H -- H_ob, edgeinto Eqn. 3, we obtain the CB band edge position results relative to water H2/H 20 level: Ecedge -Aedge .In Table 2.5, we compare the computed results with experimental data in a pH = 1 electrolyte from Refs. [105-106]. Note that our system is a 127H 20 + H30 system, so it is comparable to the pH = 1 electrolyte in terms of H+ concentration. Fig. 2.5 is plotted from the data in Table 2.5 and shows more directly the relationship between computed Ecedge -Aedge and experimental data. Test semiconductor Ecedge -A edge (eV) -exVNHE (experimental, eV) TiO 2 W0 3 CdS ZnSe GaAs GaP -0.01 -0.54 1.27 1.60 1.07 1.29 0.00 -0.20 1.50 1.50 0.80 1.10 Table 2.5 Computational results of Ecedge - Aedge and comparison with experimental data. The experimental data is translated from VNHE, the value reference to Normal Hydrogen Electrode (NHE), to Ecedge - Aedge by using ECedge - Aedge = -e x VNHE - -2.0 -- Computation Experiment -1.0 ---C : 0 -0.5H2 0/H o.0.0-- ----------------------------------------- 0)00. 0.5- ........-...................... S1.00 2 H2 0/0 2 1.5 1 1IIII TiO2 W03 CdS ZnSe GaAs GaP Figure 2.5 CB band edge level results referenced to the NHE. Blue lines are computational results by the method developed in this chapter. Red lines are experimental data from Refs. [105-106]. Two dotted lines indicate the H2/H2 0 and 02/1H20 levels in water. 2.4 Discussion Ecbulk - H,,,, A,, - H, scuk-emi -bulk W0 3 (eV) 2.35 Ecedge _ bulk H semi _ edge - H, Ecg, --A., ,, Hsot _edge IEedge -3.39 -0.70 Table 2.6 Computational results of ,, sHot Abulk - Aedge -0.34 Aedge using GGA+U for W0 3. From both Fig. 2.5 and Table 2.5, we see that our computational results are consistent with experimental data. The W0 3 system shows the largest error. To test whether this error is related to the d-character of W0 3 's CB, we repeat the computations for W0 3 using the GGA+U approximation[ 113] with U = 2.0 for the d-orbitals of W. The result, shown in Table 2.6, indicates that Ecedge - Aedge changes from -0.54 eV to -0.34 eV after applying the +U correction and shows better agreement with the experimental value of 0.20 eV. As is well known, DFT in the GGA approximation gives large errors for band gaps. However, our results for Ecedge -Aedge in Table 2.5 give an average error of 0.19 eV. We believe that the electronic level difference is in better agreement with experiment than the band gap primarily due to two reasons. One is that the computational error for band gaps comes from both CB and VB computations while our approach does not involve VB computation, so that our results do not have the error from computing VB. The other reason is that, in our approach, we are computing the energy difference between CB and LUMO, two unoccupied energy levels. They are both typically underestimated in semilocal functional[ 114]. Therefore, error cancellation may occur in their difference. Our approach can be generalized to also compute the VB band edge position relative to 0 2/H2 0 level in water. However, this may not be necessary if one has an accurate way of computing the band gap of the semiconductor, for example, using the GW approximation[ 115], hybrid or screened hybrid functional[116-120] or the A-sol[87] method. We can then determine the VB from the CB band edge position and the band gap. We also demonstrate here that the relative band edge position at a semiconductor-water interface cannot be computed by the vacuum reference approach. We take GaP as an example. By using the same approach as in Section 2.3.3, we respectively compute the Hartree potential difference at the GaP-vacuum surface and the water-vacuum surface, and denoted them as Hsemi _ edge subtracting them, we obtain - H Vacuum edge (Hsemi -edge - and Hsol Ho, edge edge HVacuum )Vacuum _approach , edge in Table 2.7. By the Hartree potential difference at the GaP-water interface by the vacuum reference method. The result is -4.18 eV (See Table 2.7). The directly computed value of Hsemiedge -Hsoledge for the GaP- water interfacial system is -3.49 eV (See Table 2.4). The discrepancy of the two results indicates that the vacuum reference approach is not valid. Hsemi GaP (eV) edge - HVacuum _ edge -7.82 Table 2.7 Result of Hsa_ edge -H Vacuum - edge -3.64 Hsemi edge - Hs, edge (Hsemi- edge - Hsl edge ) Vacuum _ approach -4.18 by the vacuum common reference approach. 2.5 Conclusions In this chapter, we present a method for computing CB band edge positions relative to the water H2/H2 0 level. The method is computationally efficient since it only involves DFT calculations with a semi-local functional. The average error, over the six compounds tested, is 0.19 eV, which makes this method useful for predicting and designing photocatalyst materials. This method and an accurate band gap DFT computation method together may provide improved knowledge of the energy levels and band gap for any photocatalyst material. Materials that are designed base on this knowledge will need little bias voltage and have a suitable band gap for photocatalysis of water-splitting. Moreover, for an arbitrary photocatalyst material, this method can tell us how large the external bias voltage should be applied in order to activate hydrogen evolution. This information is ~40~ both an important reference for experimentalists and a clue for evaluating the stability in the electrolyte of the materials. ~ 41~ Chapter III. First principles high throughput screening of oxynitrides for water-splitting photocatalysts In this chapter, we present the first principles high throughput screening system that we have constructed to search for new water-splitting photocatalysts. We use the system to screen through ~3000 nitrides and oxynitrides. Most of the known photocatalysts materials in the screened chemical space are reproduced. In addition, sixteen new materials are suggested by the screening approach as promising photocatalysts, including three binary nitrides, two ternary oxynitrides and eleven quaternary oxynitrides[71]. 3.1 Introduction In this chapter, we screen compounds using high throughput computational methods by focusing on three significant properties of water-splitting photocatalysts: (1) the crystal structure and its thermodynamic phase stability (versus competing solids and gases); (2) the band gap; (3) the conduction band (CB) and valence band (VB) edge positions relative to the H2/H2 0 and 0 2/H2 0 levels in solution. For each property, a first principles computational method has been developed which has a low enough computational cost but an adequate accuracy so that it can be used in a high throughput search. By integrating them, we thus design a three-tier high throughput screening system as following: (a) a phase stability screening to eliminate candidate compounds which are not stable enough to be synthesizable; (b) a band gap screening to eliminate all candidates with a too large or too small band gap; (c) a screening of band edge position in aqueous ~42~ environment to eliminate candidates whose CB or VB position are not suitable for watersplitting. The details of the screening system are introduced in the Section 3.2. Recently, Refs. [70, 72] computationally screened perovskite metal oxides and identified some new candidates for photocatalysts. The major differences between the screening approach in those papers and in this thesis are the following: (a) the authors in Refs. [70, 72] predicted the CB and VB positions by empirically estimating the middle of the gap using electronegativity of the atoms while we compute them directly from first principles in an aqueous environment; (b) they mainly focused on perovskite metal oxides while we consider a wider range of structures and different chemical spaces. As the reasons we mentioned in Section 1.2, we choose oxynitrides as the major chemical space to implement the screening system in this work. We screened 2948 different candidate compounds including 68 binary nitrides, 1503 ternary oxynitrides and 1377 quaternary oxynitrides. Our algorithm picked out most of the known water-splitting photocatalysts and also found sixteen new promising candidates. Some new candidates are existing materials from the Inorganic Crystal Structure Database (ICSD)[121] but have not been reported as photocatalysts yet. And some candidates are unknown compounds which are predicted by our compound prediction tool[122]. The detailed results are shown in Section 3.3. 3.2 Construction of the screening system Fig. 3.1 illustrates the high throughput screening approach in this chapter. All ~43~ computations are based on DFT[55-56] and are performed with PAW[89] potentials using the plane-wave code VASP[107-108]. For computations in step 1 and 2 in Fig. 3.1, we use the PBE[90] GGA+U[113] exchange-correlation functional unless specified otherwise, with all parameters as in Ref [84]. For computations in step 3, we use PBE GGA with all parameters as described in Chapter II as well as in Ref. [11]. SBinary 0. Candidate gen nitrides ICSD database Ternary oxy-nitridesC Quaternary oxy-nitrides 1. Phase stability s F2.Band gap screen 3s New compounds prediction DFT GGA+U, Phase stability prediction tool DFT GGA based A-sol method DFT GGA based 3-step-method for band edge position in aqueous environment Promising candidates for further,study Figure 3.1 High throughput screening approach for water-splitting photocatalysts 3.2.0 Generation of the candidates In this step, we generate the candidate compounds for the screening. As most known oxides photocatalysts contain d'0 or do cations[l, 3] (Fig. 1.5), we target primarily oxynitrides and only consider compounds that contains d'0 cations (Ga 3+, In 3 +, Ge 4 +, Sn4 , Sb 5 *, and Bi5 +) or do cations (Ti4 , Zr +, Hf4 , V'+, Nb5 *, Ta5 +, Cr6 , Mo6, W6+, Sc 3 +, and Y3+). ~44~ One of the most complete databases of experimentally observed compounds is the ICSD[121]. However, there are very few oxynitrides available in the ICSD. For example, there are only 25 ternary oxynitrides and 118 quaternary oxynitrides which contain d10 or do cations in the ICSD. Therefore, we use compound and structure prediction tools[122] to identify possible novel compounds. Since the oxynitrides space has not been as exhaustively searched with experiments as other chemistries (e.g. oxides), it is likely that there are a large number of novel compounds to be found. We used a previously developed approach based on ionic substitutions to propose new likely ternary and quaternary oxynitrides[122]. This approach uses information about the substitution probability of ions --- obtained by datamining all known crystalline compounds -- to come up with suggestions for novel compounds. To generate novel ternary oxynitrides (M-O-N, with M being a do or d10 cation), we used the set of all known binary ionic compounds as a starting point. Using the substitution algorithm from Ref. [122], we evaluated the likelihood that substituting the cation in each compound by M and the anion in the compound by a mixture of 0 and N would lead to a new stable compound. For instance, the algorithm suggested that the known Ta3N5 could have its cation Ta5 substituted by Zr4 and its anion N3 - by a mixture of 02 and N 3 -. To generate new quaternary compounds (M1-M2-O-N with MI or M2 being a do or d1o cation), we only considered a list of known ternary oxides photocatalysts in Ref. [1] as the structural framework on which to perform the substitutions. For instance, SrTiO 3 could lead to a new candidate from substituting Sr2+ by La3 , Ti4 + by Ta5 , and 02 by a mixture of N3 - and 02. according to the probabilistic model in Ref. [122]. ~-45-~ The amount of 0 and N to be substituted in each compound was determined by balancing the charge of the cations. There is however still a remaining degree of freedom in the exact ordering of 02- and N3 - on the anion sites. We enumerated the different 02- -N3 - orderings by using an algorithm similar to the one developed in Ref. [123] and selected the ones leading to the smaller cells and the larger number of N-N bonds, as the ordering of oxynitride anions has been recently shown to be driven by this factor[124]. For each candidate compound, we computed with DFT all selected orderings and only considered the one with lowest energy. It is worth noting that each possible compound also had to pass the stability screen in its relevant composition space (step 1 in Fig. 3.1) in order to be considered further. Besides the oxynitrides, we included all binary nitrides M-N with M being a transition metal or semi-metal cation from the ICSD into the screening as well. Thus, in sum, we prepared 3 batches of candidate compounds, binary nitrides (M-N), ternary oxynitrides (M-0-N) and quaternary oxynitrides (Ml -M2-0-N). 3.2.1 Phase stability screening Phase stability is an essential component of high-throughput materials discovery as new proposed candidates need to be stable enough to be synthesizable. To assess if a compound is stable at zero K, we compared its energy versus the energy of other phases or their linear combinations. This can be technically achieved through the convex hull construction[86]. Not only does the convex hull construction indicate if a compound is stable versus competing phases but this construction can be used to assess how unstable a ~46~ compound is. Therefore, we define the instability energy AH, in meV/atom, as the negative of the decomposition reaction energy to the stable phases. Stable compounds have an instability energy AH equal to zero and the larger the instability energy AH, the less stable the material is. We performed this stability analysis for all compounds considered in this work. The possible competing phases were mainly obtained from the ICSD[121]. More details on the parameters used for the computations can be found in Ref [84]. In addition, we used a recently developed scheme to mix GGA and GGA+U computations[125] as oxides and oxynitrides computations are usually performed with GGA+U while all nitrides have been computed with GGA. In this screening, we eliminated all compounds with an instability energy AH larger than 36 meV/atom. We obtained this threshold energy by doing a brief analysis of the instability energy of compounds in the ICSD. We find that more than 80% of the ICSD compounds have an "instability energy AH" less than 36 meV/atom. Since the "ICSD compounds" have, in principle, all been synthesized, we consider this threshold to be reasonable to find compounds that can be made. 3.2.2 Band gap screening In this step, we compute the band gap of the remaining candidates and eliminate those with unsuitable band gaps. Since the band gap computed from Kohn-Sham levels is usually lower than the experimental band gap by 30% ~ 100%[87], this approach cannot ~47~ be used for band gap screening. Alternatively, we use the A-sol method[87] to determine the gaps. The A-sol method is motivated by the dielectric screening properties of the homogeneous electron gas, and determines the fundamental gap from DFT total energies of systems with an electron or a hole added within the screening radius of the material. Unreliable Kohn-Sham levels are not involved in the determination of the gap, and gross underestimation of band gaps is avoided. When tested across a large number of compounds with diverse chemistries, the A-sol method gives a mean absolute error of 0.2 eV for the gap[87]. In addition, the method requires three DFT total energy computations so it is acceptable in terms of computational cost. More detailed information is available in Ref. [87]. In this screening step, we eliminate all candidates with band gaps lower than 1.3 eV or higher than 3.6 eV. The theoretical lower limit of the band gap for a water-splitting photocatalyst is 1.23 eV[1-3] but an over-potential of 0.25 eV or more is usually required[38, 126-127]. Therefore, the lowest possible band gap in practice is around 1.5 eV. We further take the mean error of the A-sol method, around 0.2 eV, into account and finally set the lower threshold to 1.3 eV. The upper threshold is more flexible. For visible light absorption, 2.7 eV could be a good upper limit. However, we extend the upper limit to 3.6 eV to also capture any interesting oxynitride materials that absorb outside visible light region. 3.2.3 Screening of band edge positions in aqueous environment In this step, we compute the CB and VB band edge positions in aqueous environment and ~-48-~ compare them with the H2/H20 and 02/1120 levels in water. We used the earlier developed three-step method[I1]. Since the method has been presented in detail in Chapter II, we do not reiterate it here. When tested on six typical photocatalysts, the method gives a mean absolute error of 0.19 eV for the CB position relative to the H2/H20 level in solution. It is worth to note that, once we obtain the CB position relative to H2/H 20 level in water, it is straight forward to obtain the VB position relative to 0 2/H20 level since the band gap has been computed in the previous step. --_1 0.7 V --------------- H2/H20 0.7 V -. 0 2/H 0 2 Figure 3.2 Allowed CB and VB positions. The red shaded area represents the allowed CB positions while the blue shaded area represents the allowed VB position. The photocatalytic water-splitting process is energetically favorable only if the CB is higher than the H2/H20 level and VB is lower than the 02/H20 level. If this is not the case, then an external bias voltage is required to shift the bands to the right positions. However, applying an external bias voltage increases the complexity of the device and more importantly requires energy input to the device thus reducing the efficiency. As a result, too large a bias voltage should be avoided. We set the threshold of the allowed bias voltage as 0.7 V and eliminate all candidates whose CB is 0.7 eV lower than the ~49~ HJH 2 0 level or whose VB is 0.7 eV higher than the 0 2/H 20 level in water as shown in Fig. 3.2. 3.3 Results of the screening 3.3.1 Binary nitrides 4 3.5 3 S2.5 CU 2- 2 1.5 CU 0.5 Binary nitrides Figure 3.3 Binary metal nitrides from the ISCD with a calculated band gap between 1.5 eV to 3.6 eV. The blue bars are computational band gaps while the red bars are experimental band gaps (if available). The experimental band gaps are collected from Refs. [3, 128-136]. We screened 68 different binary nitrides M-N where M consists of all transition metal elements and semi-metal elements. Since all of the candidate compounds are obtained from the ICSD database, we assume them to be synthesizable, so we do not present the results of their phase stability here. We find that 23 binary nitrides have a gap between 1.3 eV and 3.6 eV. These gaps and some experimental gap data (if available) are shown ~ 50 ~ in Fig. 3.3. There is generally good agreement between computational band gaps and experimental band gaps. For the 23 compounds with a suitable band gap, the band edge positions are calculated. All candidates not satisfying the band position criteria indicated in Fig. 3.2 are eliminated. The remaining candidates and their band levels are shown in Table 3.1. Material Band gap (eV) CB vs. H2/H2 0 (eV) VB vs. 0 2/H 20 (eV) GaN 3.49 -0.26 -2.63 Ge 3N4 3.59 -0.31 -2.67 Ta 3N5 2.37 0.66 (0.40 in exp) -0.49 (-0.37 in exp) Cu3N 1.99 -0.31 -1.06 AgN 3 2.68 0.48 -0.97 Zr 3N4 2.61 -0.58 -1.96 Table 3.1 Identified binary nitrides candidates. In the column of "CB vs. H2/H 20" and "VB vs. 0 2 /H 2 0", the number indicates how much the band edge is higher (more negative in the NHE reference) than the corresponding water level. Therefore, a positive number in the "CB vs. H 2 /H 2 0" column and a negative number in the "VB vs. 02/H20" column is the optimum case as that indicates that the CB and VB are bracketing the water redox levels and no bias voltage is needed. The first three nitrides in Table 3.1, GaN, Ge 3N4 and Ta 3N5 are known as water-splitting photocatalysts[44, 65-66]. Only for Ta3N5 have the band edge positions been experimentally measured[3], and these values are included in Table 3.1 for comparison. Cu 3N, AgN 3 , and Zr 3N4 are known compounds but have not been reported as photocatalysts yet. We plot their band positions relative to the water redox levels in Fig. 3.4. -1F H2/H2 0 --------------------------------- -- -- -- -- -------- - --------- W 02/H20 -w Z 2- 3Cu 3N AgN 3 Zr 3 N4 Figure 3.4 Band edge positions of Cu3N, AgN 3 , and Zr 3N 4 in the normal hydrogen electrode (NHE) reference. The solid blue lines indicates the CB levels and the solid red lines indicates the VB levels. Fig. 3.4 suggests that AgN 3 has its CB and VB bracketing the water redox levels, and thus may achieve water-splitting without an external bias voltage. However, its band gap, predicted to be 2.68 eV and experimentally measured as 3.5 eV, is suitable to absorb only UV light. Cu 3 N has a VB lower than the 02/H20 level in water but its CB is 0.3 IV lower than the H2/H20 level in water too. This indicates that H2 evolution cannot be photocatalyzed without applying an external bias voltage of at least 0.3V. However, the band gap of Cu 3N, predicted as 1.99 eV, is relatively small. This may provide enough optimization room to increase the CB while still retaining a reasonable gap. Similar to Cu 3N, Zr 3N 4 also has a too low CB and may need a bias voltage to photo-catalyze H 2 evolution. However, unlike Cu3N, the band gap of Zr 3N 4 , predicted as 2.61 eV, is already relatively large and provides little optimization room to increase the CB. As a result, Zr3N4 is likely to be less efficient as a visible light driven photocatalyst. Our screening of the binary nitrides reproduced three known binary photocatalysts, Ta 3N5 , GaN, and Ge 3 N4 . In addition, we identified three known compounds as new candidates: Cu3 N, AgN 3, and Zr 3N 4 . Cu 3 N has the potential to be an efficient visible light driven photocatalysts, while AgN 3 and Zr 3N4 are more likely to work under UV illumination. 3.3.2 Ternary oxynitrides In this section, we screened 1503 different ternary oxynitrides, M-0-N, where M is one of the do or d'0 cations mentioned in Section 3.2. The calculated phase stability of these compounds is shown in Fig. 3.5. As most of these compounds are computationally designed, it is not surprising that many of them are not stable as indicated by their large AH for decomposition in Fig. 3.5(a). Fig. 3.5 indicates that all ternary oxynitrides consisting of V5+, Cr6, Mo6+, or Sb5 are very unstable. This may be explained by the limited oxidation power of nitrogen gas. Support for this interpretation is the fact that there are no stable binary nitrides for V5+, Cr6, Mo +, and Sb 5 + in the ICSD. Moreover, we will show in the Section 3.4 that the required oxidizing power for these four cations are indeed the highest among all cations listed in Fig. 3.5. Fig. 3.5 also suggests that the majority of the stable and quasi-stable ternary oxynitrides are obtained with Ti4+, Zr4 *, Hft4 , Ta5 , Ga3 , and Ge4 much easier to oxidize. which are 1000 * 4-* 800E 600 400 * CU 20040 S20 ** M element in M-O-N (a) 140- - 6_ 200120* 100 E 80 F e( sa e e fF ** ~60 4-_ --- ** ------------ * ------------------ A----------------------- 0o-20 M element in M-O-N (b) Figure 3.5 Phase stability of ternary oxynitrides. Each point represents a different compound. The instability energy AH is defined in Section 3.2. Larger AH indicates a larger instability. Figure (a) presents all candidate compounds whose AH is less than 1000 meV/atom. Figure (b) is an enlarged version of Figure (a) focusing on the stable and quasi-stable candidates region. The red dashed line in Figure (b) represents the elimination criterion of this step, 36 meV/atom. All candidates above the red line were eliminated. Material Reported/ New AH (meV/atom) Band gap (eV) CB vs. H2/H2 0 (eV) VB vs. 0 2 /H2 0 TaON Reported 0 2.83 0.64 (0.34 in exp) -0.97 ( -0.93 in exp) Zr 2 ON 2 Reported 0 2.57 -0.34 -1.67 Ti 30 3N2 New 31 2.37 0.22 -0.92 Zr30 3N2 New 1 3.40 1.54 -0.63 (eV) Table 3.2 Identified ternary oxynitrides candidates. The numbers in the column of "CB vs. H2/H20" and "VB vs. 0 2/H20" have the same meaning as in Table 3.1. After further screening on band gap and band edge positions, four ternary oxynitrides, TaON, Zr 2 ON 2 , Zr 30 3N 2 , and Ti3 0 3N2 are identified as photocatalysts as shown in Table 3.2. Among the four ternary oxynitrides, TaON is a well-known water-splitting photocatalysts[3]. Its band edge positions have been experimentally measured[3] and are given in Table 3.2 for comparison. Zr2 ON 2 , while not present in the ICSD, has been reported as a promising material for photo-electrochemical water-splitting[137]. And the reported bixbyite structure for this compound is the same as our prediction, showing some validity of our structure prediction approach. The remaining two materials, Ti30 3N 2 and Zr 30 3N 2 have not been reported yet and are predicted by this work. Both compounds are generated from the crystal structure of Ta 3N5 . Their crystal structures are compared in Fig. 3.6. The instability energy AH is 31 meV/atom for Ti 30 3N 2 and 1 meV/atom for Zr 30 3N2 . This suggests that both materials are likely to be synthesizable. In fact, Ti3 0 3N2 is declared to have been synthesized from a website[88]. ~55 ~ (a) Ta3N5 (b) Ti 30 3N 2 (c) Zr 30 3N2 Figure 3.6 Crystal structures of (a) Ta 3N5 , (b) Ti3 0 3 N 2 , and (c) Zr 30 3N 2 . Blue atoms are N, green atoms are 0 and red atoms are (a) Ta (b) Ti (c) Zr. Structures (b) and (c) are generated by substituting all Ta atoms in (a) for Ti atoms and Zr atoms respectively and substituting 3/5 of the N atoms in (a) for 0 atoms. However, note that the positions of the 0 atoms in (b) and (c) are not identical. The band edge position shown in Fig. 3.7 suggest that Ti3 0 3N2 is particularly interesting as its CB and VB bracket the water redox levels and its band gap, predicted as 2.37 eV, is small enough for visible light absorption. Comparing the band properties of Ti 30 3N2 with TaON, the best oxynitride photocatalyst so far[3], we find that both of them have their CB and VB bracketing the water redox levels, but the band gap of Ti 30 3N2 is expected to be smaller than the band gap of TaON (2.83 eV in our computation and 2.4 eV in experiment). Therefore, Ti 3 0 3N2 has a potential to exhibit better photocatalytic performance than TaON. Zr3 0 3N2 also has its CB and VB bracketing the water redox levels, but its predicted band gap is large (3.40 eV). However, Fig. 3.7 suggests that the large band gap of Zr30 3 N2 is mainly due to its too high CB level. Shifting the CB downwards is a relatively easy band engineering problem which can be achieved with cation doping. If its CB can be shifted to be slightly higher than the H2/H2 0 level while retaining its VB position, the band gap will be reduced to 2.0 eV, making it a promising candidate for visible light driven photocatalysts. -2-1- L-------------------------------------------------------------H2/H20 w C I z 02/H20 2H Zr 3 0 3 N 2 Ti 3 0 3 N 2 Figure 3.7 Band edge position of Ti3O3N2 , and Zr 30 3N2 in the normal hydrogen electrode (NHE) reference. The solid blue lines indicate the CB levels and the solid red lines indicate the VB levels. In this section, we identified four materials, TaON, Zr2ON 2, Ti 30 3N 2, and Zr3 0 3 N2 as promising candidates for photocatalysts. TaON and Zr2 ON 2 are known photocatalysts and reproducing them from our screening system shows the validity of the approach. More importantly, we identified two new materials, Ti 3 0 3 N 2 and Zr30 3N 2 . Ti3 0 3 N 2 shows a very promising band gap and band edge positions, and has a potential to be a better visible-light driven photocatalysts than TaON. Zr3 0 3N2 is predicted to be a good photocatalysts under UV illumination, and may also be visible light driven with some CB engineering. In addition, Zr 3 0 3N 2 and Ti 3 0 3 N 2 have both the Ta 3N5 structure, and thus solid solutions of these three materials are likely to be synthesizable, creating a large chemical space in which these materials can be optimized. Further studies on the Ta3N 5:Ti 3O3N2 solid solution as a water-splitting photocatalyst will be presented in Chapter IV. 3.3.3 Quaternary oxynitrides In this section, we screened 1377 quaternary oxynitrides, Ml-M2-0-N, where MI or M2 is do or d10 metal cations as described in Section 3.2. Fig. 3.8 shows the combination of Ml-M2 elements for which we find compounds with AH less than 0.2 eV/atom. We mentioned in Fig. 3.5 that some do or d10 cations such as V 5+, Cr6 , Mo 6 +, and Sb 5 do not have stable corresponding ternary oxynitrides. In Fig. 3.8, we found that Cr6 , Mo 6, and Sb5 do have some stable corresponding quaternary oxynitrides. This suggests that the required oxidizing power for these cations may decrease by adding a second metal cation into the system. It may also be noted that, for some do or d'0 cations which have stable corresponding ternary oxynitrides in Fig. 3.5, such as Zr4 * and Ga3 +, we have not found any corresponding quaternary oxynitrides in Fig. 3.8. This is because that, as we mentioned in Section 3.2.0, our sampling in quaternary oxynitrides is not exhaustive but only based on a limited prototypes of ternary oxides listed in Ref. [1]. It is possible that stable Zr4 * and Ga3 quaternary oxynitrides exist but they cannot be derived from the prototypes we considered in this work. Another consequence of the limited sampling in ~58~ quaternary oxynitrides is that some known quaternary oxynitride photocatalysts (SrNbO 2N[138] for instance) were not reproduced by our screening since they could not be derived from the prototypes we considered. 200- -i ~La E 160- *La aLa *mg Sr *Ge 120 *Ni Sr 4*0 * Sr K Zr CrO *a*u* D120- *Ge IB Ba 80- ,La Ba F *Cs *La *Ca *Rb CO *Ba La ;Si *Li Ca ,Ba.Sr fB Srr pa,sr.S iu *Li Na ,Na a,Li,CsKI Si K I 0-0 M2 element in M1-M2-0-N Figure 3.8 Phase stability of quaternary oxynitrides. Each point represents the lowest energy compound containing the two specified cations (i.e. a do or d10 cation with another metal cation). We only show those pairs of cations which have at least one compound with AH less than 0.2 eV/atom. All candidates above the red dashed line (36 meV/atom) were eliminated. After screening for band gap and band edge positions, seventeen compounds are identified as promising candidates for photocatalysts as shown in Table 3.3. Six of these compounds, CaTaO 2N[67], SrTaO 2N[67], BaTaO 2N[67], LaTaO 2N[69], LaTiO 2N[68], and BaNbO 2N[139] have been reported as water-splitting photocatalysts. Experimentally measured band gaps for CaTaO 2N, SrTaO 2N, and BaTaO 2N are reported as 2.4 eV, 2,1 eV, and 1.8 eV respectively and our calculated band gaps for these three compounds, 2.53 eV, 2.26 eV and 1.90 eV respectively, agree well with them. Moreover, the calculated -59-~ band edge positions of these three compounds suggest that their VB are actually higher than the 0 2 /H2 0 level. Therefore, they are unsuitable for photo-catalyzing the 02 evolution reaction without an external bias voltage. This result may explain the experimental observation that only H2 evolution reaction is photo-catalyzed by CaTaO 2N, SrTaO 2N, and BaTaO 2N in Ref. [67]. Material Reported AH Band CB vs. H2/H20 VB /New (meV/atom) gap (eV) (eV) (eV) CaTaO 2N Reported 19 2.53 1.50 0.20 SrTaO2N Reported 0 2.26 1.34 0.31 BaTaO 2N Reported 0 1.90 0.97 0.30 LaTaO 2N Reported 0 1.83 0.55 -0.05 LaTiO 2N Reported 3 2.41 0.09 -1.09 BaNbO 2N Reported 0 2.03 0.59 -0.21 Ba 3Ta 20 5N 2 New 33 2.34 -0.64 -1.75 Ba 2TaO 3N New 13 2.81 -0.37 -1.95 Sr 2 NbO 3N New 0 3.15 -0.33 -2.25 Li 14 Cr 2 ON8 New 0 2.43 -0.18 -1.38 Sr 2 Ti6 O1 jN New 36 2.86 -0.15 -1.78 Ba 2Ti6 O1 jN 2 New 21 2.77 -0.11 -1.65 La 2TiO 2N 2 New 3 2.46 0.02 -1.21 Na5 MoO 4 N New 0 3.19 0.43 -1.53 Na4 WO 2 N2 New 0 2.95 0.78 -0.94 Li 5MoO 4 N New 24 2.61 1.08 -0.30 Ca5 WO 2 N4 New 0 3.26 1.71 -0.32 2 vs. 0 2/H 2 0 Table 3.3 Identified quaternary oxynitrides candidates. The numbers in the column of "CB vs. H2/120" and "VB vs. 0 2/H20" have the same meaning as in Table 3.1. ~60~ (b) Ba 3Ta 2O 5N2 (a) Sr3Ti207 Figure 3.9 Crystal structures of (a) Sr 3 Ti 2 O 7 and (b) Ba 3Ta 2O5N2 . Blue atoms are N, green atoms are 0, red atoms are (a) Ti (b) Ta, and yellow atoms are (a) Sr (b) Ba. Structure (b) is generated by substitution of all Sr atoms in (a) for Ba atoms, all Ti atoms in (a) for Ta atoms, and 2/7 of the 0 atoms in (a) for N atoms. 4 q :1 04 0 0 (a) K 2Ti6 O1 3 (b) Sr 2Ti6 0 1 jN 2 I~< 01 / &I-O '0 (c) Ba 2Ti6O1 1N 2 Figure 3.10 Crystal structures of (a) K2 Ti 60 1 3 , (b) Sr 2 Ti6 0 1 jN 2 , and (c) Ba 2 Ti6 0 1 jN 2 . Blue atoms are N, green atoms are 0, red atoms are Ti, and yellow atoms are (a) K (b) Sr (c) Ba. Structures (b) and (c) are generated by substitution of all K atoms in (a) for Sr atoms and Ba atoms respectively, and 2/13 of the 0 atoms in (a) for N atoms. ~61~ 47J (a) Sr 2 SnO4 (b) La 2 TiO2N2 Figure 3.11 Crystal structures of (a) Sr 2 SnO 4 and (b) La 2TiO 2N2. Blue atoms are N, green atoms are 0, red atoms are (a) Sn (b) Ti, and yellow atoms are (a) Sr (b) La. Structure (b) is generated by substitution of all Sr atoms in (a) for La atoms, all Sn atoms in (a) for Ti atoms, and 1/2 of the 0 atoms in (a) for N atoms. (a) Na5MoO 4N (b) Li5 MoO 4N Figure 3.12 Crystal structures of (a) Na5 MoO 4N and (b) Li5 MoO 4N. Blue atoms are N, green atoms are 0, red atoms are Mo, and yellow atoms are (a) Na (b) Li. Structure (b) is generated by substitution of all Na atoms in (a) for Li atoms. The remaining eleven materials, Ba 3Ta2 0 5N 2 , Ba 2TaO 3N, Sr 2Nb0 3N, Li 14 Cr2 ON 8 , Sr 2 Ti6 0 1 N 2, Ba 2Ti6 0 11 N 2, La 2TiO 2N 2, Na5MoO 4N, Na4 WO 2N 2 , Li5 MoO 4N, and Ca 5 WO 2N 4 have not been reported as water-splitting photocatalysts yet. Li14 Cr2 ON8 , Na 5Mo0 4N, Na4 WO 2 N 2 , and Ca5 WO 2N4 can be found in the ICSD with a known crystal structure. Ba 2 Ta0 3N and Sr 2NbO 3N are not in the ICSD but have been synthesized in Ref. [140]. Their structures are both declared as a K2NiF 4 structure. This agrees with our computation as we derived both materials from a Sr2Sn0 ~62~ 4 structure prototype, which is closely related to K 2NiF 4 . The other five candidates, Ba 3Ta 2 O 5N 2 , Sr 2 Ti6 O1 IN2 , Ba 2Ti6 O1 IN 2, La 2 TiO 2N2, and Li 5MoO 4 N are predicted by us. Ba 3Ta 20 5N2 is derived from Sr 3 Ti2 O 7 . Sr 2 Ti6 0 1 N 2 and Ba 2Ti6 O11 N 2 are derived from K 2Ti 6O 13 . La 2TiO 2N2 is derived from Sr 2 SnO 4. Li 5 MoO 4 N comes from a substitution of Li for Na in Na5 MoO 4 N. We compared their crystal structures in Fig. 3.9, Fig. 3.10, Fig. 3.11, and Fig. 3.12 respectively. The instability energy AH for Ba 3Ta 2 O 5N 2 , Sr 2 Ti6 O 11 N 2 , Ba 2Ti6 O1lN 2, La 2 TiO 2N 2 , and Li 5MoO 4 N are 33 meV/atom, 35 meV/atom, 21 meV/atom, 3 meV/atom and 24 meV/atom respectively. The band edge positions for these eleven interesting quaternary compounds are shown in Fig. 3.13, which suggests that, among them, La 2TiO 2N2 and Li 5 MoO 4 N have the best band properties for a visible light driven photocatalyst. For both, the CB and VB are bracketing the water redox levels, and their band gap is predicted as 2.46 eV and 2.61 eV respectively. The CB and VB of Na4WO 2N2 and Ca 5WO 2N4 also bracket the water redox level, but their band gaps are too large for visible light absorption. However, similarly to Zr 3 0 3N 2 , their large band gaps are mainly due to a too high CB level. Hence, if their CB levels can be shifted downwards by cation doping or solid solution, they may still become promising for visible-light driven photocatalysis. In contrast, Ba 3Ta 2 0 5N2 and Li1 4 Cr 2 ON 8 have small enough band gaps for visible light absorption, but their CB levels are lower than the H2/H 20 level, indicating that either an external bias voltage or CB engineering is needed to achieve water-splitting. It is worth noting that the CB level of Li1 4 Cr 2 ON8 is only 0.18V lower than the H2/H2 0 level, indicating that the bias voltage required is small. The remaining five materials, Ba 2 TaO3N, Sr2NbO 3N, Sr 2Ti6 O1 jN 2 , Ba 2 Ti6 O1 IN 2 , and Na 5MoO 4 N are promising candidates for photocatalysts under UV ~63~ illumination. H2/H20 Cr--------------------------------------------------------------------------- 0 /H O w 3k Ba 3Ta2 0 5N2 Ba 2TaO 3N Sr2NbO 3N L 14Cr 2ON 8 Sr2Ti O1 1N2 Ba2Ti 61 N2 -2k 1k r w I z I H --- .--_ .--._ _ _.. _.. _ _.. _ - ---_. _ -_ __ -_--_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ---_ _ __------ 2/H 22 2k 3 La 2TiO2 N2 Na 5MoO 4N Na 4WO 2 N2 Li5 MoO 4 N Ca WO2N4 5 Figure 3.13 Band edge position of Ba 3Ta 2O 5N 2 , Ba 2TaO3N, Sr2NbO 3N, Li1 4Cr 2ON8, Sr 2 Ti6 O1 N 2 , Ba 2Ti6 O1 N 2, La 2TiO 2N 2, Na 5 MoO 4N, Na4 WO 2N 2 , Li5 MoO 4N, and Ca 5 WO2N 4 in the normal hydrogen electrode (NHE) reference. The solid blue lines indicates the CB levels and the solid red lines indicates the VB levels. ~64~ In this section, we identified seventeen materials, listed in Table 3.3, as promising candidates for photocatalysts. CaTaO 2N, SrTaO 2N, BaTaO 2N, LaTaO 2N, LaTiO 2N, and BaNaO 2N have been reported as water-splitting photocatalysts while the other eleven materials, Ba 3Ta 20 5N 2, Ba 2 TaO 3N, Sr2NbO 3N, Li 1 4 Cr 2 ON8, Sr 2 Ti6 O 11 N 2 , Ba 2Ti6 O 11N 2 , La 2TiO 2N 2 , Na5 MoO 4N, Na 4 WO 2 N 2 , Li5 MoO 4N, and Ca5 WO 2N 4 are newly identified in this thesis. Among these eleven materials, La 2 TiO 2N2 and Li 5 MoO 4 N have the most promising band gap and band edge positions. Similar to Ti 3 0 3N 2 , they have the potential to be better water-splitting photocatalysts than TaON. Na4 WO 2N2 and Ca 5WO 2N4 have a band gap too large for visible light absorption and can work under UV light illumination. However, with some CB engineering, they may still be promising visible light driven photocatalysts. Ba 3Ta 2O 5N2 and Li 14 Cr 2 ON 8 have a good band gap for visible light absorption, but they need an external bias voltage to photo-catalyze the H2 evolution reaction. In addition, Ba 2TaO 3N, Sr 2NbO 3N and La 2TiO 2N2 have the same crystal structure (Sr 2SnO 4 structure). Therefore, the solid solution of these three materials are also likely to be synthesizable and are likely to be promising candidates for watersplitting photocatalysts as well. 3.4 Discussion We demonstrated in this chapter a high throughput computational screening for the design of novel water-splitting photocatalysts. Compounds are screened on phase stability, band gap, and band edge positions in aqueous environment. Eleven known photocatalysts are reproduced and sixteen new candidate photocatalysts are proposed by our screening, indicating the validity of the approach. However, as we mentioned in ~65~ Section 1.3, there are other important properties not considered in our screening. The aqueous stability of a material is important for the commercial exploitation of a photocatalyst. The standard tool to estimate it is the Pourbaix diagram[141]. However, Pourbaix diagrams have only been determined for elements in equilibrium with water. To assess the stability of more complex materials in water, one could use a recently developed method which enables the construction of Pourbaix diagrams almost entirely from first principles[142]. The method treats the Pourbaix diagram as a phase stability diagram for a material in equilibrium with various aqueous species. Our screening does not consider kinetic properties either, but they also affect the performance of photocatalysts. For instance, one major issue for hematite Fe 2O3, a promising visible light driven photocatalyst, is its poor charge carrier diffusion[143]. Defect related properties are also not considered in this work. Defect formation energy and defect concentration are closely related to the charge carrier recombination rate and lifetime, thus having an effect on the efficiency of the device. Dopability suggests to which extent the material can be engineered for a given dopant and also indicates how strong the innate P-N junction field could be designed to help the electron-hole separation. For these properties, the ab-initio predictions are often too expensive to be included in a high throughput screening system, but they could be investigated specifically for the new candidates. Another limitation is that, in this work, the properties are predicted under dark condition. However, a photocatalyst works in an illuminated environment, and thus some of its properties such as band edge positions may change accordingly. ~66~ 0 0 .E 1 bi 1 N Ni N N Figure 3.14 Estimate of the oxidizing power required to create some cations. As we mentioned in Section 3.3.2, Fig. 3.5 suggests that V5*, Cr *, Mo6 *, and Sb5 * do not form any stable ternary oxynitrides. We believe this is because the required oxidizing power for these cations is too large to achieve with Nitrogen. Fig. 3.14 illustrates this. For each cation in Fig. 3.14, we analyzed the thermodynamic phase stability of the binary oxide with the cation under different oxygen chemical potentials and thus evaluated the minimum oxygen chemical potential required to form a stable binary oxide with the cation. A higher minimum chemical potential indicates a larger oxidizing power required for the cation. While equivalent data for nitrides is not available, the oxide data in Fig. 3.14 should be representative of the relative oxidation strength needed. Fig. 3.14 shows that, V5*, Cr6 *, Mo6 +, and Sb 5 * require the largest oxidizing power among these cations. Moreover, these four cations do not have corresponding binary nitrides in the ICSD, supporting the observation we made in Fig. 3.5. It is worth noting that, we found stable quaternary oxynitrides with Cr6 *, Mo6 *, and Sb5 * in Fig. 3.8, indicating that adding a ~67~ second cation may reduce the oxidation power required. I I I I I I 0.25 N I I 0.2 E - 0.15 0.1 0.05 B g 0 0.5 1 1.5 2 2.5 3 3.5 4 Band gap reduction from binary oxide to ternay oxynitrides (eV) Figure 3.15 A histogram of band gap reduction from do and d'0 binary oxides to ternary oxynitrides. 4 * +Average band gap reduction Standard deviation of the band gap red uctior I /* 0 It \ \ II -02 2 CL CM -a 0 Cie Wq CPV 10? 4e SN / Figure 3.16 The blue line is the band gap reduction for each cation. The red line is the Standard deviation of the band gap reduction for each cation. The reason to use nitrogen instead of oxygen as the anion was to increase the valence band energy and lower the band gap. It is therefore interesting to investigate the extent of the band gap reduction as one goes from oxide to oxynitride. To study this question, we take the lowest energy structure for each ternary oxynitride that we generated in Section 3.3.2, and compare its computed band gap to the experimentally measured gap of the binary oxide with the same cation. We collected the experimental gaps from Refs. [1, 3, 144-156]. For example, for Ta5 , we compared the band gap of TaON, Ta 30 6N, Ta 4 O 7N2 , Ta 4 ON 6 , and Ta8 O1 jN 6 to the experimental measured band gap of Ta 20 5 . The reasons for using experimentally measured gaps but not computed band gaps as the binary oxide references are that (1) binary oxides are generally well studied in experiment, and thus their measured gaps are accessible and reliable; (2) we used the GGA+U based A-sol method[87] for computing the band gaps, and the U value is typically different between oxynitrides and oxides, so even if we compare the computed gaps of oxynitrides to the computed gaps of binary oxides, the comparison would not be consistent as it is based on different U parameters. A histogram of the band gap reduction for all generated ternary oxynitride compounds is shown in Fig. 3.15. Note that we only generated ternary oxynitride containing do or d10 cations. The mean band reduction is around 1.8 eV. This verifies that by introducing N into an oxide, the band gap can be significantly reduced. It is worth noting that the range of the band reduction is fairly large, from 0 eV to 4 eV. This large range could be due to two reasons. The first reason may be that the effect depends substantially on the cation. The second reason may be that, even for the same cation, introducing different amount of Nitrogen into the system may lead to different band gap reductions. To better demonstrate these two factors, we plot the band gap ~69~ reduction for each cation in Fig. 3.16. We see clearly that the mean band gap reduction is generally different for different cations, and this difference is on the order of 1 eV. In the meantime, the standard deviation of the band gap reduction for a given cation is sometimes also on the order of 1 eV. This observation indicates that the two facts mentioned above both influence the band gap reduction. 2 .5 0 . 1 1 1 1 0.3 0.4 0.5 0.6 1 I 0.8 0.9 -Average band gap reduction -Standard deviation of the band gap reduction 2 1.510. t. 0. 1 0. 2 3N/(3N+20) 0.7 1 Figure 3.17 The blue line is the average band reduction as a function of N/O ratio. The red line is the standard deviation of the band gap reduction at given N/O ratio. We look further into the effect of the amount of Nitrogen on the band gap reduction. It is possible that a change of the N/O ratio gives a different O2p and N2p weight in the valence band, thus leading to a different band gap reduction. Alternatively, a different amount of Nitrogen may lead to different crystal structures and thus to a different band gap reduction. Fig. 3.17 shows the average band gap reduction and its standard deviation as a function of N/O ratio. We observe that the average band gap reduction does not change much with the N/O ratio. From a very small ratio of N/O (left end of the blue line in Fig. ~70~ 3.17) to a very large ratio of N/O (right end of the blue line in Fig. 3.17), the average band gap reduction changes about 0.4 eV. The rest of the difference in band gap reduction, which is on the order of 1 eV, comes from crystal structure change. As an extension of this observation, we would like to point out that adding more Nitrogen into an oxide system does not necessarily lead to a larger band gap reduction as one might assume, because this statement does not consider the effect of the possible crystal structure change. For instance, our computation suggests that the band gap of Nb 4 0 7N 2 is less than that of Nb 40N6 (1.52 eV vs. 1.85 eV). 3.5 Conclusions In this chapter, we present a high throughput first principle approach to search for new water-splitting photocatalysts, and applied it to oxynitrides and some nitrides. Most of the known photocatalysts materials in the screened chemical space are reproduced, proving the validity of the approach. In addition, sixteen new materials are suggested as promising photocatalysts, including three binary nitrides, two ternary oxynitrides and eleven quaternary oxynitrides. They have been either synthesized experimentally or predicted, by our approach, to be synthesizable. Because of their predicted band gap and band edge positions, Ti3O3 N 2 , La 2TiO 2N 2 and Li5 MoO 4 N are particularly promising as visible light driven photocatalysts. In addition, with some further CB engineering or a small bias voltage, Cu 3N, Zr 30 3N 2 , Ba 3 Ta 2 0 5N 2 , Li 1 4 Cr 2 ON 8 , Na4 WO 2 N 2 , and Ca 5 WO 2N 4 also have the potential to be good visible light driven photocatalysts. The remaining seven materials, AgN 3 , Zr 3N4 , Ba 2TaO3N, Sr2NbO 3N, Sr 2Ti6 O 11N2, Ba 2Ti6 0 11N2 , and Na5MoO 4N are candidates for photocatalysts that may work under UV ~71 ~ illumination. In addition, based on our screening result, the solid solutions of Ti3 0 3N2 , Zr 30 3N 2 , and Ta3N5 and the solid solutions of Ba 2TaO 3N, Sr 2NbO 3N and La 2TiO 2N 2 may be synthesizable and may be promising candidates for photocatalysts too. ~72~ Chapter IV. First principles study on Ta3 N5 :Ti3 O3N2 solid solution as a water-splitting photocatalyst In this chapter, we propose the Ta 3N 5 :Ti 3O 3N 2 solid solution as a promising watersplitting photocatalyst. Ta 3N5 is a known visible light driven water-splitting photocatalyst while Ti 30 3 N 2 is a newly predicted compound which is proposed by the high throughput screening in Chapter III as a promising water-splitting photocatalyst. Using first principles computations, we study the phase stability, band gap, and band edge positions of the solid solution. The results suggest that the solid solution can likely be synthesized, and has a band gap lower than both its end members. The minimal band gap may be around 2.0 eV for a composition around 50%:50%, indicating that good efficiency under solar illumination may be achieved. In addition, the CB and VB of the solid solution are predicted to be bracketing the water redox levels, so the photocatalysis process is energetically favorable and bias voltage may not be required. 4.1 Introduction As we mentioned in Section 1.1, in order to achieve a ECE level above -10% under solar illumination, which is considered to be the minimum for commercialization by US Department of Energy[34], a band gap below or around 2.0 eV is necessary for a single material water-splitting system[38-39]. Although more than 130 inorganic materials have been demonstrated to exhibit photocatalytic performance for water-splitting[1], not many of them have a band gap in this range. Over 80% of the currently known water-splitting ~73~ photocatalysts are oxides[1]. The smallest band gap of these oxide compounds, which enable water-splitting without a bias voltage, is 2.3 eV[1] in the NiO/RuO 2-Ni:InTaO4 system[157-158]. It is difficult to achieve a lower band gap for oxide photocatalysts, since their valence band (VB) is typically dominated by 0 2p states which are about 3 eV lower than the H2/H 20 level[3]. Some (oxy)nitride materials have been recently reported as low band gap (below or around 2.0 eV) photocatalysts. Examples include Ta 3N5 (2.1 eV)[64], SrTaO 2N (2.1 eV)[67], BaTaO 2N (1.9 eV)[67], LaTiO 2N (2.0 eV)[68], Cao.25 LaO. 7 5TiO 2 .2 5No.75 (2.0 eV)[68], LaTaO 2N (2.0 eV)[69] and CaNbO 2N (1.9 eV)[69]. Among them, Ta 3N5 exhibits the best photocatalytic performance. It activates both the H2 and 02 evolution reaction without bias voltage, while SrTaO 2N, BaTaO2N and LaTaO 2N fail to activate the 02 evolution reaction[3]. When tested under the same condition, Ta 3N5 demonstrated about a 10 times faster 02 evolution rate and at least a similar H2 evolution rate compared to CaO. 2 5LaO.7 5TiO2.25 NO. 7 5, LaTiO 2N and CaNbO 2N[3]. However, the ECE of all these materials is still far from the commercialization target. Therefore, it is of interest to suggest new low band gap materials that have better potential as water-splitting photocatalysts. Computational studies[2, 70-73, 159] have been conducted to identify new photocatalyst materials. Especially in Chapter III, we identified sixteen known or computationally predicted (oxy)nitrides as promising water-splitting photocatalyst candidates using a first principles high throughput computational screening[71]. Among these candidates, Ti3 0 3N2, a newly predicted compound, is particularly interesting. Computational results - 74 - indicate that it can likely be synthesized[71]. In fact, a recent report on a website[88] states to have synthesized it. Its CB and VB are predicted to be bracketing the water redox levels[71], indicating that it may activate both the H2 and 02 evolution reaction without any bias voltage. The potential issue of the material may be its band gap, which is predicted to be 2.37 eV in Chapter III[71]. Although this band gap is smaller than many known (oxy)nitride photocatalysts and most of the known oxide photocatalysts, it is still not small enough to potentially achieve a commercially viable ECE. (a) Ta 3N5 (b) Ti 30 3 N2 Figure 4.1 Crystal structures of (a) Ta3N5 and (b) Ti3 0 3N2 . The structure of Ta 3N5 is obtained from ISCD [121] while the structure of Ti 30 3N2 was proposed in Chapter III[71]. Blue atoms are N, green atoms are 0, and Red atoms are Ta in (a) and Ti in (b). Ti 3 0 3N2 and Ta3N5 have the same orthorhombic crystal structure (see Fig. 4.1) with a small volume difference. The experimental lattice parameters of Ta 3N5 are a = 3.89 A, b = 10.21 A, and c = 10.26 A[160], while the lattice parameters of Ti 30 3N2 are predicted to be a = 3.86, b = 9.89, c = 9.91[71]. Thus, it may be possible to create a solid solution of Ta 3N5 and Ti 30 3N2 . As we mentioned in Section 1.2, due to band bowing effects[62-63], solid solutions can have a lower band gap than both of their end members. Therefore, it is possible that, by forming a solid solution of Ta3N5 and Ti3 0 3 N2 , one may find materials with a lower band gap than both the end members. ~75~ In this chapter, we report on the results of a first principles study of the solid solution of Ta 3N5 and Ti 30 3N 2 . We evaluate the mixing enthalpy of the solid solution at zero K to determine its phase stability. We compute the band gap of the solid solution and compare it with the band gaps of the end members. We also investigate whether the CB and VB edge positions of the solid solution, like the end members, bracket the water redox levels. The detailed methods and results of these computations are presented in Section 4.2 and 4.3. 4.2 Methods The solid solution consisting of xTa 3N5 and (1-x)Ti 3O3N2 is denoted as Ta 3 xTi 3. 3.033xN 2 +3 x. Unless specified otherwise, all DFT[55-56] computations in this chapter are performed with projector augmented wave (PAW)[89] potentials using the plane-wave code Vienna Ab-initio Simulation Package (VASP)[107-108] and the Perdew-BurkeErnzerhof (PBE)[90] GGA exchange-correlation functional. To study the phase stability of Ta 3 xTi3 .3xO 3 .3xN 2 +3 x, a set of DFT computations are performed to determine the mixing enthalpy of the solid solution at zero K, with a Monkhorst-Pack[109] 10 x 8 x 4 k-point grid and a plane wave energy cutoff of 500 eV. In these computations, we use a supercell containing 6 cations (Ta 5 + or Ti4') and 10 anions (02- or N3 -). We consider respectively 288, 225, 400, 225, 60 different possible cation and anion arrangements at five different compositions x = 1/6, 2/6, 3/6, 4/6, 5/6. Full geometric (cell shape and cell size) and ionic DFT relaxations are performed for all these configurations and also for the two end members. The mixing enthalpy of every ~76~ configuration AH?'f (x, n) is determined by Eqn. 4.1 and is normalized per cation. Here (x, n) stands for the nth configuration at the composition x. AHx (x,n) = [EDFT(x,n) - xEDFT (Ta 6 N10) - (1 - x)EDFT(Ti60 6 N 4 )]/6 (4.1) To determine the band gaps, we primarily used the A-sol method[87] which has been introduced in Section 3.2.2 and has been widely used in Chapter III. When tested across a large number of compounds with diverse chemistries, the A-sol method gives a mean absolute error of 0.2 eV for the gap[87]. In addition, the band gaps of a few (oxy)nitride compounds computed by this method in Chapter III agree well with experiments[71]. As a verification of the band gap results from the A-sol method, the band gaps of two end members and the solid solution at selected compositions are also computed using the Heyd-Scuseria-Ernzerhof functional[1 19, 161] (HSE06) with a F-centered 7 x 7 x 3 kpoint grid and a plane wave cutoff of 520 eV. Compared to A-sol, HSE06 is more accurate in predicting band gaps for typical semiconductors, but has larger errors for transition-metal compounds[87]. In addition, HSE06 is relatively more computationally expensive than the A-sol method. Therefore, we choose to primarily use the A-sol method for determining band gaps in this work. To compute the CB and VB band edge positions in aqueous environment and compare them with the H2/H20 and 0 2/H20 levels in water, we used the three-step method[ 11] which has been developed in Chapter II and has been widely used in Chapter III. The method gives a mean absolute error of 0.19 eV for the band edge positions when tested ~77~ on six typical photocatalysts in Chapter II[11], and also shows a good agreement with experimental values when applied to determine the band edge positions of a few (oxy)nitride systems in Chapter III[71]. 4.3 Results 0.15** 0 * 0.1 - * 0.1- >0.05a)* -0.1 0 1/6 2/6 3/6 4/6 xinTa 3xTi3-3xO3-3xN2+3x 5/6 1 Figure 4.2 AH,'i (x, n) of the solid solution at compositions x = 1/6, 2/6, 3/6, 4/6, 5/6. Each blue dot represents a different configuration of the solid solution. The mixing enthalpy, AH T (x, n) (eq. 1), is shown in Fig. 4.2. We observe that, there are a few configurations that have a negative enthalpy of mixing for each of the five compositions x = 1/6, 2/6, 3/6, 4/6, 5/6. Especially at compositions x = 1/6, 2/6, 3/6, 4/6, the lowest energy configurations have an enthalpy of mixing well below zero. This indicates that the ordered solid solution phase is likely dominant at zero K for all these compositions. At elevated temperatures, the increasing configurational entropy effect will further stabilize the solid solution phase. As a result, we believe that the solid solution ~78~ Ta3xTi3..3x3.3xN2+3x can likely be synthesized and remain stable for all these compositions. 2.4 2.3\,2.37x+2.37(1 2 1 10 1/6 -x)-0.831 *x*(1 -x) 2/6 3/6 4/6 5/6 1 xi a3x Ti3-3x 0 3-xN23 Figure 4.3 The band gap results of the solid solution Ta3xTi3-3x03-3xN2+3x. Blue dots represent the band gaps determined by the A-sol method. The red dash line represents E. (x) from Eqn. 4.2. E(x) = (1 - x)E2(0) + xE()) - bx(14 - x) (4.2) To determine the band gap of the solid solution, we use the A-sol method[87] to calculate the gap of the configuration with lowest energy at each composition. The band gap of both Ta3N5 and Ti303N2 determined by the same method is 2.37 eV[71], and the computed gap of Ta3N5 is in reasonable agreement with the experimental measurement, 2.1 eV[64]. We find that the band gap in the solid solution is indeed lower than the gaps of both the end members. We also notice that the band gap vs. composition x exhibits a convex shape. This is commonly known as the downward band bowing effect[62-63] and is usually described by Eqn. 4.2. In the equation, E, (O) and Eg(1) represent the band ~79~ gaps of the two end members, while bg represents the magnitude of the band bowing. In this case, Eg(0) = Eg(l) = 2.37 eV and b. = 0.831 eV may best fit the results from the A-sol method, and the corresponding Eg (x) is shown (as the red dash line) in Fig. 4.3. At composition x = 0.5, Eg (x) is minimized with a value of 2.16 eV, which is ~0.2 eV lower than the band gap of both the end members. It is worth to note that, the experimental band gap of Ta 3N5 is actually 2.1 eV instead of 2.37 eV. Therefore, we may expect the minimal band gap of the solid solution in practice to be lower than 2.16 eV. In fact, if we substitute into Eqn. 4.2 the experimental band gap of Ta 3N5 (i.e. Eg (1) = 2.1 eV), we find that, in a wide region of compositions 3/6 5 x 5/6, Eg(x) is close to 2.0 eV. In conclusion, the solid solution of Ta3N5 and Ti30 3N2 exhibits a band gap lower than the band gaps of both the end members, and the minimal band gap would be reached near composition x = 0.5 with a likely value close to 2.0 eV. Besides band gap, band edge positions are also important for water-splitting photocatalysts since the photocatalysis process is energetically favorable only if the CB and VB bracket the water redox levels. Otherwise, a bias voltage is necessary. Fig. 4.4 shows the results of the band edge positions vs. the water redox levels determined by the three-step method reported in Ref. [11]. For the two end members, Ta 3N5 and Ti30 3 N2 , the band edge positions have been computed in Ref. [71] by the same method, and are reproduced in this work. In addition, the band edge positions of Ta 3N5 have been measured experimentally[3] and the results are shown in Fig. 4.4. The computational results are in a good agreement with the experimental results. Furthermore, the band edge position of the solid solution at composition x = 0.5, i.e. Tai. 5Tii. 5 01.5N 3 .5 , is computed ~ 80 ~ and is also shown in Fig. 4.4. This solid solution corresponds to the state with minimal band gap in Fig. 4.3. Fig. 4.4 shows that both the end members and the solid solution have their CB and VB bracketing the water redox level. This indicates that no bias voltage is required for either of them. -2-1-0.66 -0.22 V -0.19V V V -. -0.4 V wfCo> 0 ----------------------------------------------------------------H /HO0 W 2 2 H2 /H 2 0 Z 0 2 /H 0 1 .15V 2- Ti3 0 3 N2 Ta 1.96V 1.71 V 5 Ti1.501. 5 N3 .5 Ta 3 N5 (com) 1.70 V Ta 3 N5 (exp) Figure 4.4 Band edge positions of Ti 30 3N 2 , Ta3N5 and their solid solution Taj.5Tij.501.5N3.5 versus the normal hydrogen electrode (NHE) reference. Both the computational and experimental results of Ta 3N5 are shown in the figure. The solid blue lines represent the CB levels and the solid red lines represent the VB levels. The number beside each line shows the value of the band edge position. 4.4 Discussion There are many possible cation and anion arrangements in the solid solution Ta3xTi 3-3x033xN2+3x. It is interesting to recognize the patterns which lead to low energy configurations. We mainly focus on pair patterns in this analysis. We observe the first and second nearest neighbor cation-cation, anion-anion and cation-anion pairs in the top 10% lowest energy configurations at each composition. In these configurations, we find that (1) for cationanion pairs, Ti-O and Ta-N are preferred rather than Ti-N and Ta-0. This is likely due to the minimization of the columbic energy. (2) For anion-anion pairs, 0 and N like to pair with the same type of anion rather than forming O-N. This agrees with the observation in Ref. [124], that the anion ordering of oxynitrides is driven to have more N-N bonds. (3) For the cation-cation pairs, Ti-Ta is slightly more favorable than Ti-Ti and Ta-Ta. These findings indicate that the low temperature ordering is likely driven by the minimization of the electrostatic energy. Although size effects, in this case likely between N3 ~and 02, tend to add a positive contribution to the enthalpy of mixing[162], the electrostatic effect is still dominant and leads to an overall negative enthalpy of mixing[163] in this system. Fig. 4.3 in Section 4.3 shows that the band gap of the solid solution has a convex shape, i.e. a downward band bowing effect[62-63]. This effect leads to the important finding that the solid solution Ta3xTi 3-3x0 3 .3xN 2 + 3x may achieve lower band gaps than both of its end members. Therefore, it is interesting to understand the mechanism of this phenomenon. The band edge position result in Fig. 4.4 shows that the solid solution Ta1 .5Tii. 50 1 .5N 3 .5 has its CB position very close to that of Ti30 3N2 and its VB position between those of Ti3 0 3N2 and Ta3N5 . This indicates that the solid solution is likely to have a CB dominated by Ti3d orbitals and a VB mainly dominated by N2p orbitals. To confirm this, we computed the density of states (DOS) diagrams of the solid solution Ta. 5 Til.501. 5N 3.5 using respectively the PBE-GGA functional[90] with a F-centered 10 x 8 x 4 k-point grid and HSE06[119, 161] with a F-centered 7 x 7 x 3 k-point grid. The GGA DOS diagram is shown in Fig. 4.5. We find that the solid solution indeed has its CB dominated ~82~ by the orbitals from Ti atoms and has its VB dominated by the orbitals from N atoms. The HSE06 DOS diagram, though is not shown here, gives the same trends. We further confirm that these dominant orbitals are mainly Ti3d and N2 p.It is known that Ti3d orbitals are lower in energy than Ta5d orbitals[164] which dominate the CB of Ta 3N5 , and N2 p orbitals are higher in energy than 02p orbitals[3] which partially contribute to the VB of Ti 3 0 3N2 . Therefore, being dominated by Ti3d orbitals and N2p orbitals at CB and VB, the solid solution exhibits a lower band gap than the interpolation of the band gaps of its end members. This explains the mechanism of the downward band bowing effect. -N -o -Ti -Ta .4-0 4- 0 Z%) -1.5 -1 -0.5 I I 0 0.5 Energy (eV) 1 1.5 2 Figure 4.5 Computed DOS of the solid solution Ta1 .5 Tii.5 O 1.5 N3 .5 by PBE-GGA functional. The DOS contributed by Ta, Ti, 0, N are colored as cyan, green, red and blue respectively. To further support the result of downward band bowing, the band gaps of Tal. 5 Ti. 5O 1.5N 3 .5, Ta 3N5 and Ti30 3N 2 have been also computed using HSE06[119, 161]. The HSE results indicate that all three materials have an indirect band gap, and the band - 83 ~ gap values are compared with the A-sol results in Table 4.1. We find that, comparing to the experimental value, the A-sol method overestimates the band gap of Ta 3N5 while HSE06 underestimates it by roughly the same magnitude, 0.2 ~ 0.3 eV. In spite of this, both methods show that the solid solution has a lower band gap than both of the end members. In fact, if we compute the magnitude of the downward band bowing at the composition x = 0.5 by Eqn. 4.3, both methods give a AEg of ~0.2 eV. Therefore, the results of band gap reduction from these two methods are consistent. Ta 3N5 (exp 2.1 eV) Taj.5Ti1 .501.5 N3 .5 Ti 30 3N2 A-sol method 2.37 eV 2.15 eV 2.37 eV HSE06 1.81 eV 1.71 eV 2.04 eV Table 4.1 Computed band gaps of Ta 3N5 , Tai.5 Tii.5 01.5 N3 .5 and Ti 30 3N 2 by the A-sol method and HSE06. AEg = 0.5Eg(Ta 3 Ns) + O.5Eg(Ti 3O3 N2 ) - Eg(Tai.s Tii.S01.5 N3 .s) In this work, we propose the solid solution Ta 3xTi 3 .3xO 3 .3xN 2 + 3x (4.3) as a promising water- splitting photocatalyst based on the computational phase diagram, band gap and band edge positions. However, some aspects which are not considered in this work may be also significant for the photocatalytic performance. For instance, we have not considered the aqueous stability even though it is important for the commercial exploitation of photocatalyst. This is mainly due to the fact that handling such a problem by first principles computations is still challenging, though several studies have recently proposed approaches to predict aqueous stability[142, 165]. While computational results are not available, relevant information from experiments can still be collected. For ~84~ instance, the aqueous stability of Ta3N5 has been tested in experiments. The material is stable under a pH > 7[64], indicating that it can work in both pure water and sea water. Ti3 0 3 N2 has not been tested so far, but oxynitrides are usually considered to be more stable in water than nitrides[3]. Kinetic properties of the solid solution are not computed in this work, but they may affect the rate of both the H2 and 02 evolution reaction, which is an important aspect of photocatalytic performance. However, in most cases, the reaction rate is not only determined by the photocatalyst itself but is also significantly affected by the co-catalysts, materials that can lower the kinetic barrier of the water-splitting reaction at the interface. For instance, the rate of H2 evolution reaction becomes ~100 times faster when 5 wt% Ru is added to TaON[43], and in the ZnO:GaN solid solution, both H2 and 02 evolution changed from negligible to clearly observable when 5 wt% RuO 2 was present[44]. 4.5 Conclusions In this chapter, we propose the solid solution Ta3,Ti3.3x03.3xN2+3x as promising water- splitting photocatalysts. By first principles computations, we study the phase stability, band gap, and band edge positions of the system. The solid solution is very likely to be synthesizable. In addition, the band gap of the solid solution is lower than the band gap of both its end members. The minimal band gap is achieved at the composition x around 0.5 and may be around 2.0 eV. This indicates that the solid solution may be able to achieve high absorption of solar illumination and obtain good efficiency. We also confirm that the CB and VB of the solid solution is bracketing the water redox levels, meaning that the ~-85-~ photocatalysis process is energetically favorable and bias voltage may not be required in this system. In conclusion, all the results suggest that the Ta 3 xTi 3 -3xO3-3xN 2 + 3 x solid solution has a good potential to achieve water-splitting photocatalysis under solar illumination with high efficiency. ~86~ Chapter V. Conclusions and future work In this thesis, we use first principles computational approaches to facilitate the design of new inorganic water-splitting photocatalysts. We have developed a so-called three-step method to compute the band edge positions relative to the water redox levels in solution. The method, which costs three DFT GGA computations for each semiconductor, gives an absolute error of 0.19 eV when tested on several typical photocatalysts. To our knowledge, no other first principles method for computing the same property has been reported with a better accuracy so far. We have constructed a high throughput screening system and have used it to identify sixteen new promising candidates for water-splitting photocatalysts from about 3000 (oxy)nitride compounds. Particularly promising candidates, such as Ti3 0 3 N 2 , La 2TiO 2N2 and Li 5MoO 4 N, have potentials to exhibit a better photocatalytic performance than any of the known water-splitting photocatalysts. The screening does not only propose promising candidates, but also suggests possible optimization strategies for the identified candidates. For instance, we have found in the screening that a particularly promising candidate Ti 3 0 3N2 has the same crystal structure as Ta 3N5 , a known photocatalyst with a low band gap. This implies that the performance of these two materials can likely be improved by forming the Ta 3N 5 :Ti 3O 3N 2 solid solution. Using first principles computations, we conclude that the solid solution may indeed be a promising water-splitting photocatalyst with a band gap ~0.2 eV lower than its end members. Due to this band gap reduction of ~0.2 eV, the maximum achievable efficiency increases from -5% to -10%. In the future, we may adapt the high throughput screening system and the related methodologies that are developed in this thesis for the screening of other promising chemical spaces, such as oxysulfides, and solid solutions of oxides, nitrides, and sulfides. Similar to oxynitrides, oxysulfides also have a higher VB position compared to oxides and may be stable in solution. In fact, a few oxysulfides have been experimentally demonstrated to exhibit a photocatalytic performance. The screening system in this thesis can be directly used to identify promising candidates for water-splitting photocatalysts among a large number of oxysulfide compounds. Solid solutions of oxides, nitrides, and sulfides are promising for low band gap photocatalysts as well. Because of the band bowing effect, solid solutions can have a lower band gap than both of their end members. Therefore, the band gap problem of oxide photocatalysts may be solved by forming solid solutions with nitrides, sulfides, or even with other oxides. For this screening purpose, two adjustments should be made to the original screening system: (1) a preparation step that groups all candidate end members. by their crystal structures should be added into the system; (2) the phase stability screening (Section 3.2.1), band gap screening (Section 3.2.2), and band edge positions screening (Section 3.2.3) should now consider solid solutions at different compositions. The screening system can also be generalized to identify photocatalysts for other applications. For example, photocatalysts can be used in various CO 2 reduction schemes to produce liquid fuels (Table 5.1). In this application, the required properties of the photocatalysts are similar to those of water-splitting photocatalysts, expect that (1) the band edge positions should now bracket different redox levels; (2) the selective kinetics at the photocatalyst-solution interface should now favor the CO 2 reduction reaction; (3) the p-type photocatalysts are preferred in this application. Therefore, we can, in principle, still screen the candidate materials based on their phase stability, band gap and band edge positions, and exam other required properties individually. The potential challenge comes from the candidates generation step (Section 3.2.0), since most known photocatalysts in this application are molecular (organic) materials but our algorithm for proposing new possible stable materials currently considers only inorganic compounds. Therefore, an alternative method needs to be developed to suggest a large number of possible candidate materials as potential photocatalysts for CO 2 reduction. Reaction, VS. NHE @ pH 7 Number of electrons Potential (V) CO2 + 2 H + 2 e~- HCO 2H 2 -0.61 CO2 + 2 H + 2 e~- CO + H2 0 2 -0.53 CO2 + 4 H + 4 e- HCHO + H2 0 4 -0.48 6 -0.38 8 -0.24 CO2 + 6 H + 6 e- -CH CO2 + 8 H + 8 e- -CH 3 0H + H 2 0 4 + 2 H2 0 Table. 5.1 Possible products for photochemical and photoelectrochemical CO 2 reduction. ~89~ Bibliography 1. Osterloh, F. E., Inorganic materials as catalysts for photochemical splitting of water. Chem Mater 2008, 20 (1), 3 5-54. 2. Liao, P.; Carter, E. A., New concepts and modeling strategies to design and evaluate photo-electro-catalysts based on transition metal oxides. Chemical Society Reviews 2013, 42 (6), 2401-2422. 3. Maeda, K.; Domen, K., New non-oxide photocatalysts designed for overall water splitting under visible light. JPhys Chem C 2007, 111 (22), 7851-7861. 4. Benson, E. E.; Kubiak, C. P.; Sathrum, A. J.; Smieja, J. M., Electrocatalytic and homogeneous approaches to conversion of C02 to liquid fuels. Chemical Society Reviews 2009, 38 (1), 89. 5. Meyer, T. J., Chemical Approaches to Artificial Photosynthesis. Accounts Chem Res 1989, 22 (5), 163-170. 6. Alstrum-Acevedo, J. H.; Brennaman, M. K.; Meyer, T. J., Chemical approaches to artificial photosynthesis. 2. Inorg Chem 2005, 44 (20), 6802-6827. 7. Ruttinger, W.; Dismukes, G. C., Synthetic water-oxidation catalysts for artificial photosynthetic water oxidation. Chem Rev 1997, 97 (1), 1-24. 8. Kumar, B.; Llorente, M.; Froehlich, J.; Dang, T.; Sathrum, A.; Kubiak, C. P., Photochemical and Photoelectrochemical Reduction of C02. Annu Rev Phys Chem 2012, 63, 541. 9. Gratzel, M., Photoelectrochemical cells. Nature 2001, 414 (6861), 338-344. ~90~ 10. Fujishima, A.; Zhang, X. T.; Tryk, D. A., TiO2 photocatalysis and related surface phenomena. SurfSci Rep 2008, 63 (12), 515-582. 11. Wu, Y.; Chan, M. K. Y.; Ceder, G., Prediction of semiconductor band edge positions in aqueous environments from first principles. PhysicalReview B 2011, 83 (23), 235301. 12. Sayama, K.; Mukasa, K.; Abe, R.; Abe, Y.; Arakawa, H., A new photocatalytic water splitting system under visible light irradiation mimicking a Z-scheme mechanism in photosynthesis. JPhotochPhotobio A 2002, 148 (1-3), 71-77. 13. Kato, H.; Hori, M.; Konta, R.; Shimodaira, Y.; Kudo, A., Construction of Z- scheme type heterogeneous photocatalysis systems for water splitting into H-2 and 0-2 under visible light irradiation. Chem Lett 2004, 33 (10), 1348-1349. 14. Abe, R., Recent progress on photocatalytic and photoelectrochemical water splitting under visible light irradiation. JPhotoch Photobio C 2010, 11 (4), 179-209. 15. Long, M.; Cai, W.; Cai, J.; Zhou, B.; Chai, X.; Wu, Y., Efficient photocatalytic degradation of phenol over Co304/BiVO4 composite under visible light irradiation. J Phys Chem B 2006, 110 (41), 20211-6. 16. Kim, H. G.; Borse, P. H.; Choi, W.; Lee, J. S., Photocatalytic nanodiodes for visible-light photocatalysis. Angewandte Chemie 2005, 44 (29), 4585-9. 17. Ye, F.; Ohmori, A.; Li, C., Formation of p-n junction by plasma spraying technique to enhance the photocatalytic activity of TiO2. Journal of Materials Science 2004, 39 (1), 353-355. 18. Chen, Y.; Crittenden, J. C.; Hackney, S.; Sutter, L.; Hand, D. W., Preparation of a novel TiO2-based p-n junction nanotube photocatalyst. Environmental science & technology 2005, 39 (5), 1201-8. 19. Liu, Z.; Zhao, Z.-G.; Miyauchi, M., Efficient Visible Light Active CaFe2O4/WO3Based Composite Photocatalysts: Effect of Interfacial Modification. The Journalof Physical Chemistry C 2009, 113 (39), 17132-17137. 20. Inoue, Y.; Sato, K.; Sato, K., Photovoltaic and photocatalytic behaviour of a ferroelectric semiconductor, lead strontium zirconate titanate, with a polarization axis perpendicular to the surface. Journal of the Chemical Society, Faraday Transactions 1 1989, 85 (7), 1765. 21. Dunn, S.; Stock, M., Artificial Photosynthesis - Use of a Ferroelectric Photocatalyst. MRS Proceedings2012, 1446. 22. Zhang, X.; Huo, K.; Hu, L.; Wu, Z.; Chu, P. K., Synthesis and Photocatalytic Activity of Highly Ordered TiO2 and SrTiO3/TiO2 Nanotube Arrays on Ti Substrates. Journalof the American Ceramic Society 2010, 93 (9), 2771-2778. 23. Matthews, R. W., Solar-Electric Water-Purification Using Photocatalytic Oxidation with Tio2 as a Stationary Phase. Sol Energy 1987, 38 (6), 405-413. 24. Mills, A.; Davies, R. H.; Worsley, D., Water-Purification by Semiconductor Photocatalysis. ChemicalSociety Reviews 1993, 22 (6), 417-425. 25. Fernandez, A.; Lassaletta, G.; Jimenez, V. M.; Justo, A.; GonzalezElipe, A. R.; Herrmann, J. M.; Tahiri, H.; Aitlchou, Y., Preparation and characterization of TiO2 photocatalysts supported on various rigid supports (glass, quartz and stainless steel). ~92~ Comparative studies of photocatalytic activity in water purification. Appi Catal BEnviron 1995, 7 (1-2), 49-63. 26. Shannon, M. A.; Bohn, P. W.; Elimelech, M.; Georgiadis, J. G.; Marinas, B. J.; Mayes, A. M., Science and technology for water purification in the coming decades. Nature 2008, 452 (7185), 301-3 10. 27. Lewis, N. S.; Nocera, D. G., Powering the planet: Chemical challenges in solar energy utilization. P Natl Acad Sci USA 2006, 103 (43), 15729-15735. 28. Huynh, W. U.; Dittmer, J. J.; Alivisatos, A. P., Hybrid nanorod-polymer solar cells. Science 2002, 295 (5564), 2425-2427. 29. Bach, U.; Lupo, D.; Comte, P.; Moser, J. E.; Weissortel, F.; Salbeck, J.; Spreitzer, H.; Gratzel, M., Solid-state dye-sensitized mesoporous TiO2 solar cells with high photonto-electron conversion efficiencies. Nature 1998, 395 (6702), 583-585. 30. Green, M. A.; Emery, K.; King, D. L.; Igari, S.; Warta, W., Solar cell efficiency tables - (Version 26). ProgPhotovoltaics2005, 13 (5), 3 87-392. 31. Frank, A. J.; Kopidakis, N.; van de Lagemaat, J., Electrons in nanostructured TiO2 solar cells: transport, recombination and photovoltaic properties. Coordin Chem Rev 2004, 248 (13-14), 1165-1179. 32. Heller, A., Conversion of Sunlight into Electrical-Power and Photoassisted Electrolysis of Water in Photoelectrochemical Cells. Accounts Chem Res 1981, 14 (5), 154-162. 33. Law, M.; Greene, L. E.; Johnson, J. C.; Saykally, R.; Yang, P. D., Nanowire dye- sensitized solar cells. Nat Mater 2005, 4 (6), 455-459. ~93~ 34. Nowotny, J.; Sorrell, C. C.; Bak, T.; Sheppard, L. R., Solar-hydrogen: Unresolved problems in solid-state science. Sol Energy 2005, 78 (5), 593-602. 35. Fujishima, A.; Honda, K., Electrochemical Photolysis of Water at a Semiconductor Electrode. Nature 1972, 238 (5358), 37. 36. Fujishima, A.; Honda, K., Studies on Photosensitive Electrode Reactions .3. Electrochemical Evidence for Mechanism of Primary Stage of Photosynthesis. B Chem Soc Jpn 1971, 44 (4), 1148-&. 37. Goetzberger, A.; Hebling, C.; Schock, H. W., Photovoltaic materials, history, status and outlook. Mat Sci Eng R 2003, 40 (1), 1-46. 38. Bolton, J. R.; Strickler, S. J.; Connolly, J. S., Limiting and Realizable Efficiencies of Solar Photolysis of Water. Nature 1985, 316 (6028), 495-500. 39. Murphy, A. B.; Barnes, P. R. F.; Randeniya, L. K.; Plumb, I. C.; Grey, I. E.; Home, M. D.; Glasscock, J. A., Efficiency of solar water splitting using semiconductor electrodes. Int JHydrogenEnerg 2006, 31 (14), 1999-2017. 40. Tang, J. W.; Durrant, J. R.; Klug, D. R., Mechanism of photocatalytic water splitting in TiO(2). Reaction of water with photoholes, importance of charge carrier dynamics, and evidence for four-hole chemistry. JAm Chem Soc 2008, 130 (42), 1388513891. 41. Ni, M.; Leung, M. K. H.; Leung, D. Y. C.; Sumathy, K., A review and recent developments in photocatalytic water-splitting using TiO2 for hydrogen production. Renewable and Sustainable Energy Reviews 2007, 11 (3), 401-425. 42. Linsebigler, A. L.; Lu, G. Q.; Yates, J. T., Photocatalysis on Tio2 Surfaces - Principles, Mechanisms, and Selected Results. Chem Rev 1995, 95 (3), 735-758. ~94~ 43. Hara, M.; Nunoshige, J.; Takata, T.; Kondo, J. N.; Domen, K., Unusual enhancement of H-2 evolution by Ru on TaON photocatalyst under visible light irradiation. Chem Commun 2003, (24), 3000-3001. 44. Maeda, K.; Takata, T.; Hara, M.; Saito, N.; Inoue, Y.; Kobayashi, H.; Domen, K., GaN : ZnO solid solution as a photocatalyst for visible-light-driven overall water splitting. JAm Chem Soc 2005, 127 (23), 8286-8287. 45. Du, X. S.; Li, Q. X.; Su, H. B.; Yang, J. L., Electronic and magnetic properties of V-doped anatase TiO2 from first principles. Physical Review B 2006, 74 (23). 46. Shao, G. S., Red Shift in Manganese- and Iron-Doped TiO2: A DFT+U Analysis. J Phys Chem C 2009, 113 (16), 6800-6808. 47. Weng, H. M.; Yang, X. P.; Dong, J. M.; Mizuseki, H.; Kawasaki, M.; Kawazoe, Y., Electronic structure and optical properties of the Co-doped anatase TiO2 studied from first principles. PhysicalReview B 2004, 69 (12). 48. Gai, Y. Q.; Li, J. B.; Li, S. S.; Xia, J. B.; Wei, S. H., Design of Narrow-Gap TiO2: A Passivated Codoping Approach for Enhanced Photoelectrochemical Activity. Phys Rev Lett 2009, 102 (3). 49. Yang, K. S.; Dai, Y.; Huang, B. B., Study of the Nitrogen concentration influence on N-Doped TiO2 Anatase from First-Principles Calculations. JPhys Chem C 2007, 111 (32), 12086-12090. 50. Di Valentin, C.; Pacchioni, G.; Selloni, A.; Livraghi, S.; Giamello, E., Characterization of paramagnetic species in N-doped TiO2 powders by EPR spectroscopy and DFT calculations. JPhys Chem B 2005, 109 (23), 11414-11419. ~95~ 51. Yang, K. S.; Dai, Y.; Huang, B. B., Understanding photocatalytic activity of S- and P-doped TiO2 under visible light from first-principles. J Phys Chem C 2007, 111 (51), 18985-18994. 52. Di Valentin, C.; Finazzi, E.; Pacchioni, G.; Selloni, A.; Livraghi, S.; Czoska, A. M.; Paganini, M. C.; Giamello, E., Density functional theory and electron paramagnetic resonance study on the effect of N-F codoping of TiO2. Chem Mater 2008, 20 (11), 3706-3714. 53. Long, R.; English, N. J., Band gap engineering of double-cation-impurity-doped anatase-titania for visible-light photocatalysts: a hybrid density functional theory approach. Phys Chem Chem Phys 2011, 13 (30), 13698-13703. 54. Long, R.; English, N. J., Band gap engineering of (N, Si)-codoped TiO2 from hybrid density functional theory calculations. New JPhys 2012, 14. 55. Hohenberg, P.; Kohn, W., Inhomogeneous Electron Gas. Physical Review 1964, 136 (3B), B864-B871. 56. Kohn, W.; Sham, L. J., Self-Consistent Equations Including Exchange and Correlation Effects. PhysicalReview 1965, 140 (4A), A1133-A1138. 57. Ji, G. H.; Gu, Z. B.; Lu, M. H.; Zhou, J. A.; Zhang, S. T.; Chen, Y. F., First principles calculations of N:H co-doping effect on energy gap narrowing of ZnO. Physica B 2010, 405 (24), 4948-4950. 58. Xu, Z.; Zheng, Q. R.; Su, G., Charged states and band-gap narrowing in codoped ZnO nanowires for enhanced photoelectrochemical responses: Density functional firstprinciples calculations. PhysicalReview B 2012, 85 (7). - 96~ 59. Wang, F. G.; Di Valentin, C.; Pacchioni, G., Doping of W03 for Photocatalytic Water Splitting: Hints from Density Functional Theory. J Phys Chem C 2012, 116 (16), 8901-8909. 60. Shi, W. J.; Xiong, S. J., Ab initio study on band-gap narrowing in SrTiO3 with Nb-C-Nb codoping. Physical Review B 2011, 84 (20). 61. Wei, W.; Dai, Y.; Guo, M.; Yu, L.; Jin, H.; Han, S. H.; Huang, B. B., Codoping synergistic effects in N-doped SrTiO3 for higher energy conversion efficiency. Phys Chem Chem Phys 2010, 12 (27), 7612-7619. 62. Zunger, A.; Jaffe, J. E., Structural Origin of Optical Bowing in Semiconductor Alloys. Phys Rev Lett 1983, 51 (8), 662-665. 63. Sun, R. S.; Ceder, G., Feasibility of band gap engineering of pyrite FeS2. Physical Review B 2011, 84 (24). 64. Hitoki, G.; Ishikawa, A.; Takata, T.; Kondo, J. N.; Hara, M.; Domen, K., Ta3N5 as a novel visible light-driven photocatalyst (lambda < 600 nm). Chem Lett 2002, (7), 736-737. 65. Yokoyama, D.; Hashiguchi, H.; Maeda, K.; Minegishi, T.; Takata, T.; Abe, R.; Kubota, J.; Domen, K., Ta3N5 photoanodes for water splitting prepared by sputtering. Thin Solid Films 2011, 519 (7), 2087-2092. 66. Sato, J.; Saito, N.; Yamada, Y.; Maeda, K.; Takata, T.; Kondo, J. N.; Hara, M.; Kobayashi, H.; Domen, K.; Inoue, Y., RuO2-loaded beta-Ge3N4 as a non-oxide photocatalyst for overall water splitting. JAm Chem Soc 2005, 127 (12), 4150-4151. ~97~ 67. Yamasita, D.; Takata, T.; Hara, M.; Kondo, J. N.; Domen, K., Recent progress of visible-light-driven heterogeneous photocatalysts for overall water splitting. Solid State Ionics 2004, 172 (1-4), 591-595. 68. Kasahara, A.; Nukumizu, K.; Hitoki, G.; Takata, T.; Kondo, J. N.; Hara, M.; Kobayashi, H.; Domen, K., Photoreactions on LaTiO2N under visible light irradiation. J Phys Chem A 2002, 106 (29), 6750-6753. 69. Hitoki, G.; Takata, T.; Kondo, J. N.; Hara, M.; Kobayashi, H.; Domen, K., (Oxy)nitrides as new photocatalysts for water splitting under visible light irradiation. Electrochemistry2002, 70 (6), 463-465. 70. Castelli, I. E.; Olsen, T.; Datta, S.; Landis, D. D.; Dahl, S.; Thygesen, K. S.; Jacobsen, K. W., Computational screening of perovskite metal oxides for optimal solar light capture. Energ Environ Sci 2012, 5 (2), 5814-5819. 71. Wu, Y.; Lazic, P.; Hautier, G.; Persson, K.; Ceder, G., First principles high throughput screening of oxynitrides for water-splitting photocatalysts. Energ Environ Sci 2013, 6 (1), 157-168. 72. Castelli, I. E.; Landis, D. D.; Thygesen, K. S.; Dahl, S.; Chorkendorff, I.; Jaramillo, T. F.; Jacobsen, K. W., New cubic perovskites for one- and two-photon water splitting using the computational materials repository. Energ Environ Sci 2012, 5 (10), 9034-9043. 73. Jain, A.; Castelli, I. E.; Hautier, G.; Bailey, D. H.; Jacobsen, K. W., Performance of genetic algorithms in search for water splitting perovskites. Journal of Materials Science 2013. ~98~ 74. Hautier, G.; Jain, A.; Ong, S. P.; Kang, B.; Moore, C.; Doe, R.; Ceder, G., Phosphates as Lithium-Ion Battery Cathodes: An Evaluation Based on High-Throughput ab Initio Calculations. Chem Mater 2011, 23 (15), 3495-3508. 75. Jain, A.; Hautier, G.; Moore, C.; Kang, B.; Lee, J.; Chen, H. L.; Twu, N.; Ceder, G., A Computational Investigation of Li9M3(P207)(3)(PO4)(2) (M = V, Mo) as Cathodes for Li Ion Batteries. JElectrochem Soc 2012, 159 (5), A622-A633. 76. Chen, H. L.; Hautier, G.; Jain, A.; Moore, C.; Kang, B.; Doe, R.; Wu, L. J.; Zhu, Y. M.; Tang, Y. Z.; Ceder, G., Carbonophosphates: A New Family of Cathode Materials for Li-Ion Batteries Identified Computationally. Chem Mater 2012, 24 (11), 2009-2016. 77. Hautier, G.; Jain, A.; Chen, H. L.; Moore, C.; Ong, S. P.; Ceder, G., Novel mixed polyanions lithium-ion battery cathode materials predicted by high-throughput ab initio computations. JMater Chem 2011, 21 (43), 17147-17153. 78. Wang, S. D.; Wang, Z.; Setyawan, W.; Mingo, N.; Curtarolo, S., Assessing the Thermoelectric Properties of Sintered Compounds via High-Throughput Ab-Initio Calculations. Phys Rev X 2011, 1 (2). 79. Madsen, G. K. H., Automated search for new thermoelectric materials: The case of LiZnSb. JAm Chem Soc 2006, 128 (37), 12140-12146. 80. Armiento, R.; Kozinsky, B.; Fornari, M.; Ceder, G., Screening for high- performance piezoelectrics using high-throughput density functional theory. Physical Review B 2011, 84 (1). 81. Hachmann, J.; Olivares-Amaya, R.; Atahan-Evrenk, S.; Amador-Bedolla, C.; Sanchez-Carrera, R. S.; Gold-Parker, A.; Vogt, L.; Brockway, A. M.; Aspuru-Guzik, A., The Harvard Clean Energy Project: Large-Scale Computational Screening and Design of ~99~ Organic Photovoltaics on the World Community Grid. J Phys Chem Lett 2011, 2 (17), 2241-2251. 82. Sokolov, A. N.; Atahan-Evrenk, S.; Mondal, R.; Akkerman, H. B.; Sanchez- Carrera, R. S.; Granados-Focil, S.; Schrier, J.; Mannsfeld, S. C. B.; Zoombelt, A. P.; Bao, Z. N.; Aspuru-Guzik, A., From computational discovery to experimental characterization of a high hole mobility organic crystal. Nat Commun 2011, 2. 83. O'Boyle, N. M.; Campbell, C. M.; Hutchison, G. R., Computational Design and Selection of Optimal Organic Photovoltaic Materials. J Phys Chem C 2011, 115 (32), 16200-16210. 84. Jain, A.; Hautier, G.; Moore, C. J.; Ong, S. P.; Fischer, C. C.; Mueller, T.; Persson, K. A.; Ceder, G., A high-throughput infrastructure for density functional theory calculations. Comp Mater Sci 2011, 50 (8), 2295-23 10. 85. Ceder, G.; Hautier, G.; Jain, A.; Ong, S. P., Recharging lithium battery research with first-principles methods. Mrs Bull 2011, 36 (3), 185-191. 86. Hautier, G.; Ong, S. P.; Jain, A.; Moore, C. J.; Ceder, G., Accuracy of density functional theory in predicting formation energies of ternary oxides from binary oxides and its implication on phase stability. PhysicalReview B 2012, 85 (15). 87. Chan, M. K. Y.; Ceder, G., Efficient Band Gap Prediction for Solids. Phys Rev Lett 2010, 105 (19). 88. http://www.titandioxide.ru/titan s/sc3/0176.php. 89. Bl6chl, P. E., Projector augmented-wave method. Physical Review B 1994, 50 (24), 17953-17979. ~100~ 90. Perdew, J. P.; Burke, K.; Ernzerhof, M., Generalized Gradient Approximation Made Simple. Phys Rev Lett 1996, 77 (18), 3865-3868. 91. De Angelis, F.; Fantacci, S.; Selloni, A., Alignment of the dye's molecular levels with the TiO(2) band edges in dye-sensitized solar cells: a DFT-TDDFT study. Nanotechnology 2008, 19 (42), 424002. 92. Coe, J. V.; Earhart, A. D.; Cohen, M. H.; Hoffman, G. J.; Sarkas, H. W.; Bowen, K. H., Using cluster studies to approach the electronic structure of bulk water: Reassessing the vacuum level, conduction band edge, and band gap of water. J Chem Phys 1997, 107 (16), 6023-6031. 93. Rhoderick, E. H., Metal-Semiconductor Contacts. lee Proc-I 1982, 129 (1), 1-14. 94. Van de Walle, C. G.; Neugebauer, J., Universal alignment of hydrogen levels in semiconductors, insulators and solutions. Nature 2003, 423 (6940), 626-8. 95. Franciosi, A.; VandeWalle, C. G., Heterojunction band offset engineering. Surf Sci Rep 1996, 25 (1-4), 1. 96. Majewski, J. A.; Stadele, M.; Vogl, P., Stability and band offsets of SiC/GaN, SiC/AIN, and AlN/GaN heterostructures. Mater Res Soc Symp P 1997, 449, 917-922. 97. Cheng, J.; Sprik, M., Aligning electronic energy levels at the TiO2/H20 interface. Physical Review B 2010, 82 (8). 98. Cheng, J.; Sulpizi, M.; Sprik, M., Redox potentials and pK(a) for benzoquinone from density functional theory based molecular dynamics. J Chem Phys 2009, 131 (15). 99. Shaltaf, R.; Rignanese, G. M.; Gonze, X.; Giustino, F.; Pasquarello, A., Band offsets at the Si/SiO(2) interface from many-body perturbation theory. Phys Rev Lett 2008, 100 (18). ~101~ 100. Alkauskas, A.; Broqvist, P.; Devynck, F.; Pasquarello, A., Band offsets at semiconductor-oxide interfaces from hybrid density-functional calculations. Phys Rev Lett 2008, 101 (10). 101. Mrovec, M.; Albina, J. M.; Meyer, B.; Elsasser, C., Schottky barriers at transition- metal/SrTiO3(001) interfaces. PhysicalReview B 2009, 79 (24). 102. Grossman, J. C.; Schwegler, E.; Draeger, E. W.; Gygi, F.; Galli, G., Towards an assessment of the accuracy of density functional theory for first principles simulations of water. JChem Phys 2004, 120(1), 300-311. 103. Schwegler, E.; Grossman, J. C.; Gygi, F.; Galli, G., Towards an assessment of the accuracy of density functional theory for first principles simulations of water. II. J Chem Phys 2004, 121 (11), 5400-5409. 104. Prendergast, D.; Grossman, J. C.; Galli, G., The electronic structure of liquid water within density-functional theory. J Chem Phys 2005, 123 (1). 105. Nozik, A. J.; Memming, R., Physical Chemistry of Semiconductor-Liquid Interfaces. The JournalofPhysical Chemistry 1996, 100 (31), 13061-13078. 106. Nozik, A. J., Photoelectrochemistry: Applications to Solar Energy Conversion. Annual Review of Physical Chemistry 1978, 29 (1), 189-222. 107. Kresse, G.; Furthmtiller, J., Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. PhysicalReview B 1996, 54 (16), 11169-11186. 108. Kresse, G.; Joubert, D., From ultrasoft pseudopotentials to the projector augmented-wave method. PhysicalReview B 1999, 59 (3), 1758-1775. 109. Monkhorst, H. J.; Pack, J. D., Special points for Brillouin-zone integrations. PhysicalReview B 1976, 13 (12), 5188-5192. 110. Smith, W.; Forester, T. R., DL_POLY_2.0: A general-purpose parallel molecular dynamics simulation package. JMol Graphics 1996, 14 (3), 136-141. 111. Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L., Comparison of Simple Potential Functions for Simulating Liquid Water. J Chem Phys 1983, 79 (2), 926-935. 112. Hermida-Ramon, J. M.; Karlstrom, G., Study of the hydronium ion in water. A combined quantum chemical and statistical mechanical treatment. J Mol Struc-Theochem 2004, 712 (1-3), 167-173. 113. Dudarev, S. L.; Botton, G. A.; Savrasov, S. Y.; Humphreys, C. J.; Sutton, A. P., Electron-energy-loss spectra and the structural stability of nickel oxide: An LSDA+U study. Physical Review B 1998, 57 (3), 1505-1509. 114. Komsa, H. P.; Broqvist, P.; Pasquarello, A., Alignment of defect levels and band edges through hybrid functionals: Effect of screening in the exchange term. Physical Review B 2010, 81 (20). 115. Hedin, L., New Method for Calculating 1-Particle Greens Function with Application to Electron-Gas Problem. Physical Review 1965, 139 (3A), A796. 116. Becke, A. D., Density-Functional Thermochemistry .3. The Role of Exact Exchange. J Chem Phys 1993, 98 (7), 5648-5652. 117. Lee, C. T.; Yang, W. T.; Parr, R. G., Development of the Colle-Salvetti Correlation-Energy Formula into a Functional of the Electron-Density. Physical Review B 1988, 37 (2), 785-789. 118. Perdew, J. P.; Emzerhof, M.; Burke, K., Rationale for mixing exact exchange with density functional approximations. JChem Phys 1996, 105 (22), 9982-9985. ~103~ 119. Heyd, J.; Scuseria, G. E.; Ernzerhof, M., Hybrid functionals based on a screened Coulomb potential. J Chem Phys 2003, 118 (18), 8207-8215. 120. Krukau, A. V.; Vydrov, 0. A.; Izmaylov, A. F.; Scuseria, G. E., Influence of the exchange screening parameter on the performance of screened hybrid functionals. J Chem Phys 2006, 125 (22). 121. Inorganic Crystal Structure Database Inorganic Crystal Structure Database (httW//icsd.fiz-karlsruhe.de/icsd/). 122. Hautier, G.; Fischer, C.; Ehrlacher, V.; Jain, A.; Ceder, G., Data Mined Ionic Substitutions for the Discovery of New Compounds. Inorg Chem 2011, 50 (2), 656-663. 123. Hart, G. L. W.; Forcade, R. W., Algorithm for generating derivative structures. Physical Review B 2008, 77 (22). 124. Yang, M. H.; Oro-Sole, J.; Rodgers, J. A.; Jorge, A. B.; Fuertes, A.; Attfield, J. P., Anion order in perovskite oxynitrides. Nat Chem 2011, 3 (1), 47-52. 125. Jain, A.; Hautier, G.; Ong, S. P.; Moore, C. J.; Fischer, C. C.; Persson, K. A.; Ceder, G., Formation enthalpies by mixing GGA and GGA plus U calculations. Physical Review B 2011, 84 (4). 126. Wrighton, M. S.; Ginley, D. S.; Wolczanski, P. T.; Ellis, A. B.; Morse, D. L.; Linz, A., Photoassisted Electrolysis of Water by Irradiation of a Titanium-Dioxide Electrode. P Natl Acad Sci USA 1975, 72 (4), 1518-1522. 127. Walter, M. G.; Warren, E. L.; McKone, J. R.; Boettcher, S. W.; Mi, Q. X.; Santori, E. A.; Lewis, N. S., Solar Water Splitting Cells. Chem Rev 2010, 110 (11), 6446-6473. 128. T. Lindgren; M. Larsson; Lindquist, S. E., Proc. 14th Int. Workshop on Quantum Solar Energy Conversion 2002. ~104- 129. Smith, A. R.; AL-Brithen, H. A. H.; Ingram, D. C.; Gall, D., Molecular beam epitaxy control of the structural, optical, and electronic properties of ScN(001). J Appl Phys 2001, 90 (4), 1809-1816. 130. Bergman, L.; Chen, X. B.; McIlroy, D.; Davis, R. F., Probing the AlxGal-xN spatial alloy fluctuation via UV-photoluminescence and Raman at submicron scale. Appl Phys Lett 2002, 81 (22), 4186-4188. 131. Hara, M.; Hitoki, G.; Takata, T.; Kondo, J. N.; Kobayashi, H.; Domen, K., TaON and Ta3N5 as new visible light driven photocatalysts. Catal Today 2003, 78 (1-4), 555560. 132. Bar, M.; Ahn, K. S.; Shet, S.; Yan, Y.; Weinhardt, L.; Fuchs, 0.; Blum, M.; Pookpanratana, S.; George, K.; Yang, W.; Denlinger, J. D.; Al-Jassim, M.; Heske, C., Impact of air exposure on the chemical and electronic structure of ZnO:Zn3N2 thin films. Appl Phys Lett 2009, 94 (1). 133. Becker, J. S.; Kim, E.; Gordon, R. G., Atomic layer deposition of insulating hafnium and zirconium nitrides. Chem Mater 2004, 16 (18), 3497-3501. 134. Rees, C. S.; Chaudhri, M. M., The Electrical-Conductivity of Silver Azide (Agn3) at High Hydrostatic Pressures. JPhys C Solid State 1987, 20 (26), 4097-4104. 135. Kroll, P., Hafnium nitride with thorium phosphide structure: Physical properties and an assessment of the Hf-N, Zr-N, and Ti-N phase diagrams at high pressures and temperatures. Phys Rev Lett 2003, 90 (12). 136. Chen, J. J.; Gila, B. P.; Hlad, M.; Gerger, A.; Ren, F.; Abernathy, C. R.; Pearton, S. J., Band offsets in the Sc203/GaN heterojunction system. Appl Phys Lett 2006, 88 (14). ~105~ 137. Mishima, T.; Matsuda, M.; Miyake, M., Visible-light photocatalytic properties and electronic structure of Zr-based oxynitride, Zr2ON2, derived from nitridation of ZrO2. Appl Catal a-Gen 2007, 324, 77-82. 138. Maeda, K.; Higashi, M.; Siritanaratkul, B.; Abe, R.; Domen, K., SrNbO2N as a Water-Splitting Photoanode with a Wide Visible-Light Absorption Band. JAm Chem Soc 2011, 133 (32), 12334-12337. 139. Siritanaratkul, B.; Maeda, K.; Hisatomi, T.; Domen, K., Synthesis and Photocatalytic Activity of Perovskite Niobium Oxynitrides with Wide Visible-Light Absorption Bands. Chemsuschem 2011, 4 (1), 74-78. 140. Tobias, G.; Beltran-Porter, D.; Lebedev, 0. I.; Van Tendeloo, G.; Rodriguez- Carvajal, J.; Fuertes, A., Anion ordering and defect structure in Ruddlesden-Popper strontium niobium oxynitrides. Inorg Chem 2004, 43 (25), 8010-8017. 141. Pourbaix, M., Atlas of ElectrochemicalEquilibriain Aqueous Solutions (National Association of CorrosionEngineers,Houston, Texas) (1974). 142. Persson, K. A.; Waldwick, B.; Lazic, P.; Ceder, G., Prediction of solid-aqueous equilibria: Scheme to combine first-principles calculations of solids with experimental aqueous states. PhysicalReview B 2012, 85 (23). 143. Dareedwards, M. P.; Goodenough, J. B.; Hamnett, A.; Trevellick, P. R., Electrochemistry and Photoelectrochemistry of Iron(Iii) Oxide. J Chem Soc FaradT 1 1983, 79, 2027-2041. 144. Zhai, H. J.; Li, S.; Dixon, D. A.; Wang, L. S., Probing the electronic and structural properties of chromium oxide clusters (CrO3)(n)(-) and (CrO3)(n) (n=1-5): ~106~ Photoelectron Spectroscopy and density functional calculations. J Am Chem Soc 2008, 130 (15), 5167-5177. 145. Higashiwaki, M.; Sasaki, K.; Kuramata, A.; Masui, T.; Yamakoshi, S., Gallium oxide (Ga2O3) metal-semiconductor field-effect transistors on single-crystal beta-Ga203 (010) substrates. App Phys Lett 2012, 100 (1). 146. Zyubin, A. S.; Mebel, A. M.; Lin, S. H., Optical properties of oxygen vacancies in germanium oxides: Quantum chemical modeling of photoexcitation and photoluminescence. JPhys Chem A 2007, 111 (38), 9479-9485. 147. Zhu, W. J.; Tamagawa, T.; Gibson, M.; Furukawa, T.; Ma, T. P., Effect of Al inclusion in HfO2 on the physical and electrical properties of the dielectrics. Ieee Electr Device L 2002, 23 (11), 649-651. 148. Piper, L. F. J.; DeMasi, A.; Cho, S. W.; Smith, K. E.; Fuchs, F.; Bechstedt, F.; Korber, C.; Klein, A.; Payne, D. J.; Egdell, R. G., Electronic structure of In203 from resonant x-ray emission spectroscopy. Appl Phys Lett 2009, 94 (2). 149. Mai, L. Q.; Hu, B.; Chen, W.; Qi, Y. Y.; Lao, C. S.; Yang, R. S.; Dai, Y.; Wang, Z. L., Lithiated MoO3 nanobelts with greatly improved performance for lithium batteries. Adv Mater 2007, 19 (21), 3712. 150. Kurmaev, E. Z.; Moewes, A.; Bureev, 0. G.; Nekrasov, I. A.; Cherkashenko, V. M.; Korotin, M. A.; Ederer, D. L., Electronic structure of niobium oxides. JAlloy Compd 2002, 347 (1-2), 213-218. 151. Nainani, A.; Sun, Y.; Irisawa, T.; Yuan, Z.; Kobayashi, M.; Pianetta, P.; Bennett, B. R.; Boos, J. B.; Saraswat, K. C., Device quality Sb-based compound semiconductor surface: A comparative study of chemical cleaning. JAppl Phys 2011, 109 (11). ~107~ 152. Emeline, A. V.; Lobyntseva, E. V.; Ryabchuk, V. K.; Serpone, N., Spectral dependencies of the quantum yield of photochemical processes on the surface of wideband-gap metal oxides. 2. Gas/solid system involving scandia (Sc2O3) particles. J Phys Chem B 1999, 103 (8), 1325-133 1. 153. Kilic, C.; Zunger, A., Origins of coexistence of conductivity and transparency in SnO2. Phys Rev Lett 2002, 88 (9). 154. Chakrabarti, A.; Hermann, K.; Druzinic, R.; Witko, M.; Wagner, F.; Petersen, M., Geometric and electronic structure of vanadium pentoxide: A density functional bulk and surface study. PhysicalReview B 1999, 59 (16), 10583-10590. 155. Jia, D. D.; Wang, X. J.; Yen, W. M., Delocalization, thermal ionization, and energy transfer in singly doped and codoped CaAl407 and Y203. Physical Review B 2004, 69 (23). 156. Emeline, A.; Kataeva, G. V.; Litke, A. S.; Rudakova, A. V.; Ryabchuk, V. K.; Serpone, N., Spectroscopic and photoluminescence studies of a wide band gap insulating material: Powdered and colloidal ZrO2 sols. Langmuir 1998, 14 (18), 5011-5022. 157. Zou, Z. G.; Arakawa, H., Direct water splitting into H-2 and 0-2 under visible light irradiation with a new series of mixed oxide semiconductor photocatalysts. J Photoch PhotobioA 2003, 158 (2-3), 145-162. 158. Zou, Z. G.; Ye, J. H.; Sayama, K.; Arakawa, H., Direct splitting of water under visible light irradiation with an oxide semiconductor photocatalyst. Nature 2001, 414 (6864), 625-627. 159. Kanan, D. K.; Carter, E. A., Band Gap Engineering of MnO via ZnO Alloying: A Potential New Visible-Light Photocatalyst. The Journal of Physical Chemistry C 2012, 116 (18), 9876-9887. 160. Brese, N. E.; O'Keeffe, M.; Rauch, P.; DiSalvo, F. J., Structure of Ta3N5 at 16 K by time-of-flight neutron diffraction. Acta CrystallographicaSection C CrystalStructure Communications 1991, 47 (11), 2291-2294. 161. Heyd, J.; Scuseria, G. E.; Emzerhof, M., Hybrid functionals based on a screened Coulomb potential (vol 118, pg 8207, 2003). J Chem Phys 2006, 124 (21). 162. Tepesch, P. D.; Kohan, A. E.; Garbulsky, G. D.; Ceder, G.; Coley, C.; Stokes, H. T.; Boyer, L. L.; Mehl, M. J.; Burton, B. P.; Cho, K. J.; Joannopoulos, J., A model to compute phase diagrams in oxides with empirical or first-principles energy methods and application to the solubility limits in the CaO-MgO system. Journal of the American Ceramic Society 1996, 79 (8), 2033-2040. 163. Ceder, G.; Kohan, A. F.; Aydinol, M. K.; Tepesch, P. D.; Van der Ven, A., Thermodynamics of oxides with substitutional disorder: A microscopic model and evaluation of important energy contributions. Journal of the American Ceramic Society 1998, 81 (3), 517-525. 164. Kato, H.; Kudo, A., Photocatalytic water splitting into H2 and 02 over various tantalate photocatalysts. Catal Today 2003, 78 (1-4), 561-569. 165. Chen, S.; Wang, L.-W., Thermodynamic Oxidation and Reduction Potentials of Photocatalytic Semiconductors in Aqueous Solution. Chem Mater 2012, 24 (18), 36593666. ~109~